andyreagan.com

• Found it on HN
• Just enough abstraction of software problems that it might be helpful
• E.g., first define the business context (like an insurance policy is a thing, it can have 1 insured (can that change?))
• Then you can build the software components that map to these at a high level
• And then on the finer level you can have the software do what it needs to function

• Train.py is ~200 lines of python code that define and train GPT-2 from scratch using mlx and numpy as the only dependencies. This readme will detail writing train.py from scratch. The model is trained on ~1 million characters of Shakespeare contained in input.txt, and it can be trained in around 10 minutes on a macbook to produce coherent Shakespeare-like text.
• This would be a good basis for giving a talk on how the models work.

• HN
• Fun little idea to organize your personal scripts. Title about says it all

• NFC, electronics breakdown
• Pretty neat how it works, it's a little passive chip without a processor
• Cool blog on the whole

• Amazing inspirational stuff: how to use computers to think
• New project: Dynamicland

• HN
• HN seems to call out that she has been suing apple for all kinds of things…EPA doesn't report a problem

• There is a whole conference
• Term coined by https://www.inkandswitch.com/local-first/ (interesting group)
• Blog about it using dropbox to achieve it: https://tonsky.me/blog/crdt-filesync/

• https://alexanderobenauer.com/ollos/ everything in a timeline

• Supposed to be better than copilot

• "it seems your experience with tracking "everything" already lead you to the conclusion that the value of the data lies in the role it plays in interrogating your own beliefs and mental models and improving them"
• Big list of these aggregators like exist, gyroscope: https://wiki.openhumans.org/wiki/Finding_relations_between_variables_in_time_series

• Maybe a better way to do secret handling (from HN - what's the threat model).
• The encryption key is still in a file (.env.keys)
• It moves a vulnerability on file access to one on process env access (which is worse?)

• Still can't plan

• If you (or your team) are shooting yourselves in the foot constantly, fix the gun
• Assess the trade-off you’re making between quality and pace, make sure it’s appropriate for your context
• Spending time sharpening the axe is almost always worth it
• If you can’t easily explain why something is difficult, then it’s incidental complexity, which is probably worth addressing
• Try to solve bugs one layer deeper
• Don’t underestimate the value of digging into history to investigate some bugs
• Bad code gives you feedback, perfect code doesn’t. Err on the side of writing bad code (same as quality-pace maybe)
• Make debugging easier
• When working on a team, you should usually ask the question
• Shipping cadence matters a lot. Think hard about what will get you shipping quickly and often

• Like Yahoo pipes, GIU editor with filters and flows

• HN
• Dystopian, it tries to sound like a person who read your stuff (it did…but it's not a person)

• Chile is so long, it's curved.

• Spreadsheets are a powerful programming abstraction
• I think tools like Notion and such are indeed close
• Code allows you to do anything
• My idea: blocks. Start with a document, blocks are each paragraph. Like observable cells, but markdown default.
  • Blocks > Documents > Userspace

• HN
• Well told story of running an indian chemical manufacturing business for a year

• HN
• It can run shell scripts like perl with backticks
• Syntax is more compact than Python
• But it's not install everywhere

• HN
• There was a presentation/slides on it that were fascinating
• The author

• Focus on 10x work, downplay the rest
• Build in the open
• Intergrating novel ideas may be part of that (making connections), but idk if this HN habit helps

• Quantam GPS…could replace satellites!

• Yeah, I see this…
• The only extension I have that's not from GitHub or Microsoft is Awesome Emacs Keymap. Replace it by just adding these keys to a personal dotfiles?

Backlinks: Long term outlook

• HN
• I'm not sure I buy the premise that a super-intelligence is about to come

• Maybe, maybe not. Could be good process and still give awards…

• HN
• Yeah, not great at it.
• Update: Claude 3.5 is better.

• Chart comparing the zones from different models (3-zone, 5-zone, etc)

• LLMs guiding genetic programming…interesting.

• Also Awesome ARC
• ARC is a prize for AI, with 2d puzzles that are hard even for humans to solve
• Learn a pattern from three examples of before/after, and then apply it
• Maybe the exact opposite of having a human do numerical computing (have the computers do visual processing)
• Join the competition? The constraints are pretty tight…
• The one that seems hard to program is filling in circled areas

Question - could Claude generate summaries this good of all these resources? Would reading those summaries allow me to remember less than reading the actual resources?

HN

https://simonwillison.net/2024/Jul/31/aider/

https://github.com/paul-gauthier/aider

Update <2025-04-29 Tue>: I never got Aider set up but I have been using Claude Code which is the same idea (but perhaps more expensive!).

https://alexplescan.com/posts/2024/08/10/wezterm/

WezTerm, terminal emulator. Terminals are software GUIs for devs…

I've been using this, it's…the same to me as iTerm.

https://hackerdads.wordpress.com/2017/02/22/finding-out-more-about-me-for-free/

Not a lot to take away here. Dude is trying to do some QS stuff.

https://earlyretirementnow.com/2016/12/07/the-ultimate-guide-to-safe-withdrawal-rates-part-1-intro/

https://arxiv.org/abs/2402.09799

Love this idea.

• Every company should be owned by its employees | Hacker News

• Anyone can access deleted and private repository data on GitHub | Hacker News

• Bayesian Statistics: The three cultures | Hacker News

• The New Internet | Hacker News

Ambitious post about how Tailscale is creating a new wave of internet. Not so sure about this…

https://100r.co/site/home.html

• 100R — library
  • Hundred Rabbits is a small collective exploring the failability of modern tech | Hacker News

• Neat CSS

The wet bulb: the most human of temperature measurements. We are wet bulbs!

https://tacobellquarterly.org/a-prayer-for-the-annihilation-of-every-automobile/

Amazing.

https://github.com/scottbez1/smartknob

https://www.unihertz.com/products/titan

Heard it's what fly uses, and now gitpod: https://github.com/firecracker-microvm/firecracker

I like this! Touchscreens in cars are hard to use.

https://news.ycombinator.com/item?id=42108360 https://asteriskmag.com/issues/08/the-death-and-life-of-prediction-markets-at-google

https://longevitybiomarkers.com/

cache

iOS app to track biomarkers and give you a score. Not sure where the score comes from - not a lot of details! Incorporates some interesting markers besides bloodwork:

ALL BIOMARKERS AVAILABLE IN THE APP:

Body Mass Index • Vitamin D • Blood Pressure • Orthostatic Hypotension • Peripheral Artery Status • Resting Heart Rate • VO2 max Heart Rate Variability • Body Fat • Visceral Fat • Skeletal Muscle • Strength • Flexibility • Balance • Functional Fitness • Walking Speed • Hearing Reaction Time • A1C • Albumin • Apo B • Cystatin C • IGF-1 • Chronic Inflammation

https://austinhenley.com/blog/mirrorlang.html

cache

Really cool idea - you just write the tests! Of course, the AI could get stuck, but if your tests are simple enough…it could work.

https://www.gauge.sh/blog/ai-makes-tech-debt-more-expensive

Cache: cache.

I really like this post because it's so true. If you can break down your work into small, manageable pieces this makes it possible for both humans and AI. Caution: breaking it down is often the hard part!

Backlinks: Genetic Testing

Plug in your gene data.

https://www.foundmyfitness.com/genetics

https://www.sumsar.net/blog/pandas-feels-clunky-when-coming-from-r/

Backlinks: Longevity

https://www.levels.com/blog/the-2024-levels-guide-to-metabolic-health-interventions

Backlinks: Longevity

https://www.nature.com/articles/s43587-024-00692-2

Have I just gone through one? I got so much slower.

Backlinks: Longevity

https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0310764

https://ehandbook.com/leaving-google-to-join-an-early-stage-startup-a-framework-for-making-hard-decisions-f18504ad67f4

Doesn't feel like the right time for me, but maybe it is!

With Annie Duke.

Rather than a pros and cons list, Duke suggested people make that forecast by imagining, say, how you think you’ll feel a year from now. “Where are you happier?”. Avoid Loss Aversion.

Cool ring of blogs?

Career planning. hn Make a plan so you at least know what you're aiming for, even if it's only 5 years.

Good writing on time management. Hat tip to Cal Newbort and David Allen.

HN

Deep dive, wonderful. Differentiated controls are very helpful. See Audi vent controls as a master class. Everything is a design problem.

Backlinks: Writing as thinking

and AI will chip away at the need for this skill.

A positive spin / satire on cease and desist letters. Kind of a fun idea, but the letter would totally scare someone.

HN

A play on best fucking website, pro minimal sites.

https://arstechnica.com/culture/2024/11/flour-water-salt-github-the-bread-code-is-a-sourdough-baking-framework/?utm_source=tldrnewsletter Pretty fun.

boxes as used by Adam Savage for incredible organization:

https://www.mysortimo.us/en_US/Storage-Bins-%26-Boxes/T-BOXX/T-BOXX-320/T-BOXX-320-incl-IB-Set-16-pieces-H63-S/p/000000006000015479

He built shelves that allows these to pull out.

Timeline building DSL. Inspired by markdown.

• HN: https://news.ycombinator.com/item?id=42289690 At least someone mentions that they should just use org mode :joy:

https://news.ycombinator.com/item?id=42351101 drop in replacement for make. like it! github: https://github.com/casey/just has a shell export option so you're not totally locked in.

data]]

:ID: E9813678-019E-469A-A025-CCE50C54AAFC

Cautionary tale. Took some steps noted in Hardening google dependency.

HN: https://news.ycombinator.com/item?id=42325011

cache

His blog has all sorts of great little essays. This one certainly wasn't my favorite.

I had this downloaded from Nov 5… Great post! A classic.

HN

Can't consume all the time dude - got to produce. I resonate a small amount with the time I take reading the top HN links, like this one. I do pick up valuable ideas… but it's a lot of info to process. E.g., I have some quantam computer stuff open, Google Willow, but I don't have time to understand quantam computers right now.

Interesting…

YC startup offering low cost customer ETF indexes. Finances are so complicated!

HN

Founder of redis is returning to redis after they yanked the open source license from the community. I like his post. That's not his accounting of what happened on the license, sounds like the HN community is pretty cynical?

HN I like this breakdown.

https://news.ycombinator.com/item?id=42330055 Didn't read.

Self attesting a code of ethics. I like it.

https://egoless.engineering/ cache

HN I like this approach a lot.

Author of boring technology club.

https://www.outsideonline.com/health/training-performance/2024-sweat-science-holiday-books/

Some hip stuff: Hip rotations: https://www.youtube.com/watch?v=VI2HPy3zfhY 90/90 rear foot lift: https://www.youtube.com/watch?v=K3L3UYSf86k https://www.healthline.com/health/hip-internal-rotation?utm_source=Six+Minute+Mile&utm_campaign=7caadcb616-EMAIL_CAMPAIGN_2020_07_28_03_29_COPY_01&utm_medium=email&utm_term=0_6e5b2f993e-7caadcb616-13291137&mc_cid=7caadcb616&mc_eid=3e23310d23#exercises-and-stretches

https://news.ycombinator.com/item?id=42388488

https://www.ntia.gov/sites/default/files/publications/january_2016_spectrum_wall_chart_0.pdf

Could rebuilt the turkey trot site on this so others can edit…

One dude made a new espresso machine design. Pretty simple TBH!

fitness age

Financial skills training

See when friends are in town Maybe not a problem we need an app for? Or is it an ideal use of social media? What if you don't want to see someone in particular…

from Lumpy on Strava. Amazing stuff here, apparently he died though…

Turns out the URLs come through in the email, their link removal is post-post.

https://news.ycombinator.com/item?id=42419469 Pretty interesting the amount of details that AI can get here. At some level, just unsuprising compared to what a private detective could get from your photos (a bit worse). On another level, wow they know a lot just from photos.

https://news.ycombinator.com/item?id=42417330 Tangentially related to personal knowledge management.

Didn't realize these things existed… I think our house is leaky enough.

Dumb TVs…I like the idea. But paying $190 for an Apple TV just so that Vizio isn't spying on me seems like a high price!

It will just blend in, I agree.

HN

People suggest clojure too!

Cool

HN Interesting post, basically all about information discrepancy. How can I know a van is good, I only get to try a few cars in my life!

Helpful strategies!

Interesting?

HN Pretty hilarious parody website of oil and gas selling "lives"

to get a smartphone. hn Good idea. Works if everyone does it.

without much outcome.

a calendly alternative.

• HN
• Gwern has an article about this: https://gwern.net/problem-14
• Description of portfolio management in a similar game style https://andrewcharlesjones.github.io/journal/universal-portfolios.html
• Original paper from Cover on portfolio optimization: https://isl.stanford.edu/~cover/papers/paper93.pdf

https://news.ycombinator.com/item?id=20077177 Similiar idea with images: https://news.ycombinator.com/item?id=41877482 Webpage with multiple "pages" as a single page: https://github.com/cadars/john-doe

Cool!

Lots of similar projects, like https://alexsci.com/feeds/. Love that webring: https://xn--sr8hvo.ws/

https://arcprize.org/blog/oai-o3-pub-breakthrough O3 model does well on ARC-AGI, maybe it's getting better. Generalization is part of the problem. The hard task seems to be developing just the right abstraction, which could very well be embodiment. Just like alpha zero can do chess, go, etc, and now o3 can do this, there are tasks like GSM-Symbolic that are still too hard (yet very easy for humans), also ARC-AGI-2. Just have to get the right abstraction / DSL! HN The cost axis seems like a strawman, it's interesting if we can solve these with any amount of compute that's less than brute force.

https://news.ycombinator.com/item?id=42475390

Love this idea. Sam's a hard pass, bummer.

https://www.anthropic.com/research/building-effective-agents

HN

https://news.ycombinator.com/item?id=42483895

Lean could be providing the right DSL for math, which is pretty fascinating (recall the math challenge success of some LLMs).

• City roads: https://news.ycombinator.com/item?id=42481206 Could use this to map emergency service proximity?
• Grayjay app - a reader for multimedia content. HN I should learn RSS better!

A few points: AGI conservation feels old - yes. You have to use it cautiously if you think it's "talking to an expert". It's useful already - yes! "for the most part that AI revolution is not going to be because of PHD-level LLMs" more data won't work forever "inference time scaling": https://www.aisnakeoil.com/p/is-ai-progress-slowing-down

https://news.ycombinator.com/item?id=42488983 Docs: https://twtxt.readthedocs.io/en/latest/index.html Example site: https://txt.sour.is/user/xuu/twtxt.txt Maybe nostr is where it's at https://github.com/nostr-protocol/nips/blob/master/01.md Link blogging: https://simonwillison.net/2024/Dec/22/link-blog/ Activity pub (mastodon protocol) type things (like peertube): https://news.ycombinator.com/item?id=42490313

Keep is simple. Even simpler. Good stuff! HN

Fun story, watched the youtube (skipped around a bit).

HN Can get by without actually understanding. LLM is like this.

HN

Tea bags have microplastics? From a very questionable journal. Cellulose is not microplastic, is it?!

Yes I like this idea. Should not need 600 pages for an idea.

Interesting stuff. I should try clojure! Can it go real world stuff.

Insidetracker is expensive.

https://news.ycombinator.com/item?id=42505454 Web version from Gemini: https://gistpreview.github.io/?76f03f49be58344bfa64c9d5d9f0ec29 Github: https://github.com/Jaennaet/pISSStream

(cache) I mean, yes!

Super interesting results. Definitely concerning as an endocrine disruptor. BPA in Salmon the only result that seems actionable. HN My avoidance of hot things and plastic is warranted. More glass, more cast iron.

Backlinks: Long term outlook

While these folks waste breath debating whether AI is useful, I’m going to be over here…using it.

Interesting discussion, but I didn't get any new takes.

Interesting stuff happening out there. Sounds like maybe Russia is just doing chaos.

1:60 scale model of an airplane made only with manila folder and glue. Pretty incredible how far he's taken this concept.

I agree! Maybe the GPT models will too, and this is just a honeymoon period like we had w Google.

HN Some interesting sounding talks. I'm writing this in Emacs right now!

Maybe the market out there really is so tough!

HN Apple photos sending hashes of photos off the device? Seems like a violation of privacy, but they're also doing it in the most secure way possible.

• gyroscope use to be awesome. They led the game. But they started charging more for less, almost no new integrations, and hidden pricing. I commented as such in the subreddit and was banned and DMd by the CEO. Do not recommend. It’s expensive for almost no benefits.
• exist is ok, but I wouldn’t invest too much in it. They had a chance to lead the market, but they never scaled beyond two developers and their app has grown at a slower rate. they're still going!
• Welltory: has become my go-to, but they’re starting to slip. The lack of a decent desktop dashboard is deflating and whenever I open the mobile app it just doesn’t seem to work
• IFTTT (custom): set up some custom apps that pull your data into gsheets, then import it into a database that you can upload to a custom AI if you feel comfortable doing so.
• Reflect: devs joined in
• Guava mentioned by one other as well

Hadn't heard of these, good to know they exist!

Cool self-hosted service. Other good home lab recommendations in the thread. I had a bad experience with Plex, felt like ad-ware (and I'm running this myself!). Jellyfin would have been better, perhaps.

The examples here seem insane, this is a research language. Maybe I should learn Lean!

There are leap seconds. 29 so far.

HN Very quick parse - a description of passkeys. Which seem like an improvement, maybe, for people who don't use password managers.

• (cache)
• HN

Maybe worth a read!

Ultimate Python Cheat Sheet: Practical Python For Everyday Tasks I could have written this? But they did.

Excellent list!

Random blog, inspiring directness, found it from: https://guzey.com/people/

Recently I found "Things" for my tools for thought page. I should try an RSS reader like Reeder Classic. Otherwise some good overlap. I could keep track of my own list… https://flexibits.com/fantastical seems interesting.

Good news!

Yes good points! The design is not always fully decribed by the code itself.

cache Interested in how the derived the 22-minute number. New study: https://onlinelibrary.wiley.com/journal/1971 Old study with 11 minutes: https://www.bmj.com/content/320/7226/53.1 So the new study link is just a link to a journal, I couldn't find the article there. None of the four links from the brief actually link to the article. The brief summary leads with "Professor Sanjay Agrawal" but by Googling one of the lines, I found a link to the article via google: https://onlinelibrary.wiley.com/doi/10.1111/add.16757 Actual pdf: ./static/smoking-20-minutes.pdf

https://www.rtl-sdr.com/using-a-hackrf-and-javascript-browser-app-to-perform-rolljam-replay-attacks-on-a-car/

Framework to help think about how tech is layers. Very cool blog + projects, indie web stuff.

The work of reducing tech debt, making things work. Not just writing the code that works in isolation. I do a lot of this.

HN Pretty solid summary. From HN:

I built and shipped a Swift app to the App Store, currently generating $10,200 in MRR, exclusively using LLMs.

Wow!

Yeah… Nothing too new here.

A lot of research on this actually, nothing suprising:

People with more money can afford healthier food, more healthcare and homes in safer, less-polluted neighborhoods, says Kathryn Himmelstein, a co-author of the study and a medical director at the Boston Public Health Commission. “We know the things that help us age better, and everyone’s always disappointed when you tell them,” says Andrew Scott, director of economics at the Ellison Institute of Technology in Oxford, England. “Eat less and eat better, sleep more, exercise more and spend time with friends.”

how does a technology scale down?

Really good question - when and how to use simpler technology? The web is complicated with JS and WASM and everything, to the point that building a browser is like building an operating system. This is a lot like how I'm trying to use Emacs and plain text as the foundation of my web site.

Really good guide! Start from the top, rather than the easy "independent" pieces. Envelope-math size critical parts first with simple test scripts.

You start coding at the top of the stack, letting each layer define the API of the layer below it.

• Show HN: API Parrot – Automatically Reverse Engineer HTTP APIs Kind of like using web dev tools, but more automated. Similar: Launch HN: Integuru (YC W24) – Reverse-engineer internal APIs using LLMs Low-latency integrations with platforms lacking official APIs. Cool idea too.
• Rails for everything I like this premise, but use Django! HN > Rails is awesome, and so is Django.
• 30% drop in O1-preview accuracy when Putnam problems are slightly variated Meh, not clear whether this is real. They are very good pattern matchers but everyone still wants to know whether they reason.
• One of my papers got declined today (HN link) Terrence Tao shares that not every one of his papers gets rejected. The academic journal process as he describes it is very ideal, but that may not always be the case. Good to share failures! https://www.dahosek.com/2024-in-reejctions-and-acceptances/ This guy graphs his!
• Nutrionic AI - A new iOS (only) nutrition analytics app He's building an app - work in progress. Not ready to even test yet.
• Passkey technology is elegant, but it's most definitely not usable security > I cannot bring myself to agree to any "switch to passkey" prompt from any device because I have no idea (and too tired to figure out) how and where that key will be stored, how do I deal with different devices, etc. I already have a universal solution for credentials: 1password, which is cross-platform. Yes… > Having to have multiple passkeys per site to circumvent vendor identity lock-in is one of the main problems with passkeys Yes! > The problem with any solution relying on a couple of physical devices as the sole access to your digital life is that the management and protection of those objects becomes one of the most important things in your life. Eesh, losing all digital access may not be that dire? I agree with this article. Tying access to Apple/Google and to a physical device is upgraded security for the masses but also harder tie-in to that ecosystem, it's more fragile. List of stories of account lock-out. Lots of good points here, one commenter notices that there are really two different groups of people: the techies who will use Yubikey and everyone else who will be sucked into the solution that the UX pushes them to. I've had to work around using the iOS passkeys, I bet Sam has some on her device, it's so pushy to use them. Can they actually transfer from iOS to android?? Seems like they can't - which is a big problem.

Link Pretty cool thing to do, I'm sure you'd get better at solving problems in SQL!

Some interesting advice - mid size firms and consulting can be fun.

HN Dude needs some friends or a family. Sounds lonely. Other maybe helpful advice from the hive mind.

Author writes a very basic blog. I like it! Lots on HN also appreciate these things. Let's internet like it's 2005!

Seems good when they publish false charges.

Same WFH kind of stuff we've heard. I avoided the management ladder and love it.

A lot of very interesting parts to this story! HN

Extracts contents from the android app - the actual model. Pretty fun. HN Uses a tool called frida: https://www.youtube.com/watch?v=CLpW1tZCblo

It's about how you make people feel. HN

Open postgres got taken over. It was in docker, so not a huge deal to contain. HN Love for tailscale on HN, it works great for me.

HN The author has the right idea - to be "senior dev" good at coding, it's about developing the mental models of how the code fits together. Abstracting just the right amount.

But it's not a complete test.

HN It's totally not ready yet. But this does seem to be a huge golden goose for wearable technology.

Not every day your work gets mentioned in the NYT!

But seven, too, was not destined to last. In 2016, data scientists at the University of Vermont, exploiting advances in natural language processing, plugged 1,737 stories scraped from the Project Gutenberg website into a software program in order to chart their “emotional arcs.” Their custom “hedonometer” examined each story in blocks of 10,000 words, assigning a “meaningful sentiment score” to every chunk, and allowing the researchers to express the results as a series of nifty graphs. The one for “Harry Potter and the Deathly Hallows” shows a mountain range of peaks and valleys, ending, after Voldemort is finally killed, on a steep upward incline toward a summit at happily ever after. The study made news: The world’s stories boiled down to “six basic shapes.”

Here's a copy of the article.

cache HN Solarcan, pinhole camera…very cool: https://solarcan.co.uk/product/solarcan

Backlinks: Alcohol

cache Nuanced take on research around alcohol. Definitely bad for you, but also not clear that light drinking is very harmful.

Backlinks: Alcohol

This is a review. Basically one study that shows probiotics help with cravings. Another with fecal transplantion helped with cravings. Still very early, looks like prebiotics and other non-direct interventions did not affect cravings in AUD.

Interesting thought, but it would definitely be worse. See HN.

It seems to me many people are so amazed by the fact they for once had an original thought that at this moment they stop caring whether it is actually a good one. It is crucial to maintain mental flexibility and one does that by thinking things through, killing your darlings, admitting when ideas are wrong or simply just mediocre. Only because it was me who had an idea doesn't mean I have to defend it at all cost. The idea isn't me.

Very pie-in-the-sky vision for the software. I love these kinds of big ideas and grand visions. Whether it holds up? Maybe for very simple pages? HN good discussions. It doesn't play nicely with other frameworks or custom js you have to write, an aspect of software that I think about a lot (soft edges, jumping levels of abstraction).

Top level programmer.

I foresee a world with far more specialized code, with fewer generalized packages, and more readable tests. Reusable code will continue to thrive around small robust interfaces and otherwise will be pulled apart into specialized code. Depending how well this is done, it will lead to either better software or worse software. I would expect both, with a long-term trend towards better software by the metrics that matter.

HN

Very fun little game.

$3,000 to run unlimited full size models. Cool! There is already a mini version for 250 that can run 8B models.

HN Cool tool to render 3d surfaces.

HN Two things going on - the author brings a valid point (renaming the account is in violation of the CC license for content) along with complaining after violating the rules. SO needs to just delete the account, they're not required to host it forever.

Interesting example of how to actually break an RSA key. Possible because it's too short.

First, maybe it's not real. HN points out he's only been there 1.5yr, and he says about to graduate. But if it's real, dang the university should not be threatening him like that!

Nothing new at CES…innovate or die!

App to do correlations on tracked behavior. Seems like it will be cool. So many of these apps… Testflight available: https://apps.apple.com/us/app/cora-data-driven-lifestyles/id6737651703

Meh

You must resist the urge to make your little corner of the codebase nicer than the rest of it.

Interesting take, good principles. This guy wrote the glue work post I just saw too. HN My take is that fighting to make things better is really the job.

HN link.

Due to their liquid form, SSBs are rapidly consumed and digested

Lustig, maybe a quack (who knows…): https://robertlustig.com/sugar-the-bitter-truth/

Pretty interesting - pre-training for a tabular model!

Could be interesting, based on Polar360 wearable.

Software/SDK for subscription management. Used by Ladder / OpenAI.

Sounds like good stuff. HN

After all the drama HN What a disaster

All from them just bricking devices. Some alternatives to "connect" older systems: Wiim Pro & Mini Airplay server: https://github.com/juhovh/shairplay Maybe sonos will turn it around?

Good principles. Mine: believe the error messages. hn git bisect

HN Looks like this was not approved, Snyk is sloppy or their founder has a competitor to Cursor? Hmmmm. Hard problem checking code in dependencies…

LEGO is great.

Stopping the fighting seems good. Lots of talk of genocide. Caught this link of horrific content on the original hamas attack: https://www.hamas-massacre.net/categories/rockets-striking-civilian-areas

Digital versus hard copy - hard copies do help with memory. Something different about holding the book. Digital has some advantages though too. Saw DEVONthink - another tool for thought perhaps.

They're just done. I didn't know what they are - they do investigate investing (find frand, short companies, report them). Shares this to watch: https://www.youtube.com/watch?v=h96MGcsi7GQ They're short on carvana, made a mountain on Elon's twitter takeover (both from a short and long at the right times). Roblox sounds like a disaster. hn

No shit, he doesn't need a fast mimicing agent when he doesn't need to fast! Genuinely not suprised. He does seem like he's trying to help, hard to be so genuine on those long Youtube interviews (the one with Magnus was pretty good). But you really never know, could just be selling snake oil (in the form of fancy olive oil).

Yes! Almost 5 years and going.

I signed up, it's pretty cool to see the full 5 year trends. Seeing my best streaks is cool too. I like it, but the annual price is a lot. https://www.reddit.com/r/whoop/comments/1i0fhv1/my_belated_fitiq_and_whoop_year_in_review/

I agree w the author, every time this happens it gets just a little bit worse to walk. HN Jaywalking can be safer, it's legal in NYC. Lots of other mixed feelings.

Wrong, they do change the "sub" as they should.

But with foreign labor and shipping overseas to package? Maybe some progress here.

pretty interesting take - he is the algorithm, manifested. some interesting stuff.

Okay yes I should just run tiny tiny RSS: https://tt-rss.org/. Reddit posts with images: https://www.reddit.com/r/homelab/search.rss?q=self%3Atrue&restrict_sr=on&sort=top&t=year HNRSS as a middleman to HN: https://hnrss.org/newest?points=150 This feed: https://joeyehand.com/feed_rss_created.xml > Bluesky is trying to figure out how to outsource algorithms and let you decide which to use. https://oxide-and-friends.transistor.fm/episodes/scaling-bluesky-with-paul-frazee > OPML blogrolls https://opml.org/blogroll.opml https://alexsci.com/rss-blogroll-network/discover/feed-c550c508de4e5f2500785a76c79ba383/ http://scripting.com/ https://blogroll.social/ newsletter to rss: https://kill-the-newsletter.com/ rsshub will do twitter https://docs.rsshub.app/routes/social-media#x-twitter > There still exist a couple of nitter instances which provide RSS feeds for X.

• Indieweb: https://indieweb.org/social_reader
• Congress: https://www.govinfo.gov/rss/bills.xml
• SEC: https://www.sec.gov/news/pressreleases.rss
• List: https://www.govinfo.gov/feeds
• https://minifeed.net/blogs/
• https://ooh.directory/
• https://blogscroll.com/
• https://boredreading.com/
• https://github.com/kagisearch/smallweb
• https://blog.davidv.dev/blogs-i-follow.html
• https://rss.surf/
• https://lighthouseapp.io/

Fun adversarial idea. Not sure I want to battle the bots with my own electricity and compute…

> Markov chain isn't adversarial enough

Small GPTs fighting the bigger ones?

It can work! We could use this tool to give out access codes for the wellness program. Everyone gets an access code, GRAIL can check them. But would we get the log of which ones are being checked? Yes. Grail would have to call some random ones?

Pretty fun read.

I should really check this out (again).

Turns out that a lot of people I knew posted huge life updates that I completely missed out on.

eah I regret dropping IG. I still read a lot of HN.

Seems politically timed?

I should try with a basic tiling library…

Impressive project! Company Railway is building an AWS competitor.

• GovTrack.us
• NPR
• state senator
• Substack
• hn
• Lots of good suggestions there.
• Turns out org-mode has a feed reader.

https://writepermission.com/org-blogging-unfurling-links.html They seem to have customized it very well, and leaned into writing the code!

cache HN has some good adversarial ideas He ended up adding a proof of work - cool idea.

cache Really deep take on why bees die. Take every queestion like this! hn, didn't read

love it. hn is harsher > adopt a real dog

hn kalman filters are so cool! didn't read yet though.

Apparently this 3d printer phones home to china constantly. It's possible to use via an SD card and never connect to the internet but you don't get firmware updates. Feels so much like 2d printers, I applied the firmware update from our epson printer and now it doesn't take third party ink cartridges. Recommended: Prusa MK4S. Would need to model stuff using Solid Edge or Fusion 360.

Sort of interesting. Defaults matter.

The original post on enshittification, in some sense. A twitter thread before Cory Doctorow wrote his piece. Basic take is that software has moved to a model where ownership is not like physical goods via DRM and subscriptions and closed source. hn Offline maps: https://osmand.net/ > Convenience kills How much work would it to be to not be locked in to these various ecosystems? I try pretty hard! Looking for more prescriptions. how can I have both the convience of the apple ecosystem and none of the BS he mentions? Some other things

• LibreOffice
• Steamdeck
• https://github.com/sindresorhus/awesome

Some pretty interesting takes here. Anti-tech, for sure.

Awesome list. Cache. Love so many of these.

locally hosted code assistant. I should really try getting a local model running and hooked up to Emacs.

Interesting!

First posted on show HN in 2009!

cache Good idea! Make them smaller. Implementation is a bit harder, the very slow compression seems like the real blocker. Seems like it can be an option, like how the author is using it. hn cache similar idea to switch to brotli, back in 2023.

Pebble failed, sold to fitbit, bought by Google, they just released. The future of rebble, hn Rebble is an open port of the web stuff for Pebble post failure. The original founder of Pebble is going to try making them again (see hn). Watchy interesting open source project in the space, costs $60.

I need to try UV!

From Simon Willison. Used the web interface here. Discussion on hn. Run deepseek r1 on your own hardware for $6k. Positive reviews of Aider, a bridge between APIs and help coding.

Pretty cool! Next they have to build a full size plane.

Obsidian, Anki, Things.

Pull it all down yourself using Python. Reflect. Guava. Excel.

cache With all the flood the zone news on trump, and expected outrage, was curious to read a pro argument. Maybe some interesting points about lowering standards…a fundamental call for meritocracy. There could definitely be a safety element in military too (I wouldn't want to go into a burning building with someone totally unprepared physically). Tough questions. Military fitness standards are interesting, here is the RAND study. They actually got rid of the leg tuck option, here are the standards. I should got for a test, have most of the stuff. Need a 10lb med ball to throw.

I was hoping to see someone suggest to just run a forum. I'd love to do this for Shutesbury. Front porch forum: could I run this in Shutesbury?

• https://joinmobilizon.org/en/
• https://groups.io/

One thing I really like about FPF is that it's daily.

Great podcast! Life is uncertain. Trying to control everything is recipe for stress. The interviewee has a podcast: https://www.powercorruptspodcast.com/episodes

Plane crash. Horrific. Can see the tracks online, very interesting.

Why DeepSeek had to be open-source (and why it won't defeat OpenAI) HN Meh, it's not really open source if you can't build it!! Arc Analysis of DeepSeek's R1-Zero and R1 HN They don't beat it like o3 did, they're more like the o1. OpenAI mad they trained on output Meh, OpenAI trained on our outputs!

Love the article.

Jesus. Y’all, if you get an email from me it will be signed with HUGS, LOVE, FIGHT THE PATRIARCHY, DOWN WITH POWDERED GRAVY or SORRY I SUCK SO MUCH. It will be filled with typos and rambling parentheticals and apologies for answering several months too late. This is how you know it’s me and not a robot.

Also

It’s like my whole computer is a toddler screaming “LET ME DO IT!” every time I try to create something.

Yes!

Up to 30% off!! That's a LOT of noise.

@article{stubbs2025predictive,
  title={Predictive equation helps estimate misreporting of energy intakes in dietary surveys},
  author={Stubbs, R James and Hopkins, Mark},
  journal={Nature Food},
  pages={1--2},
  year={2025},
  publisher={Nature Publishing Group UK London}
}

PDF

Cool! 10 days until valentines day… One of the commenters does the exact opposite, lol.

Growth mindset. Steve Magness is great! And people are complex: not all good or evil. Can vary on the situation and change. Accept who you are in all situations: radical acceptance (though he doesn't call it that).

We need to revive it! Have a thread w The Weather Company behind wunderground. Should contact others…

Feb 22-23, AI Agents. Of course I'd want to build a health one. Prizes are pretty wimpy - I'd want funding from a16z to start the company!

Interesting that Apple's "default" software can take over more space. Kind of like Google, except they don't tend to abandon things so much.

Explains some differences. Need local security (maslow) to extend more broadly.

From Derek Lowe, what a wild time. Basically I don't think they are legally allowed to control funding, but the chaos is real. HN Some interesting takes, and different opinions/experiences on what DEI actually is doing. Some saying that it really is seeking equality of opportunity, but plently of examples for equality of outcomes. Open letter from Tom McAndrew on this Forecasting of flu is going to be worse!

Good luck to them!

Some takes about how this is self-serving, but I would certainly love if all my TV subscriptions, spotify, whoop, etc all only charged me if I used it that month!

Code forces humans to be precise. That’s good—computers need precision. But it also forces humans to think like machines. … Try writing a tax code in chat messages. You can’t.

Yes, he's obviously right. But I'd also say that it's about writing good prompts. You just have to give it all of the information, concisely. HN

80% of my job is already trying to figure out what's actually needed Building the wrong thing fast is about as useful as not building anything at all.

Yes!

Can follow his guide here.

Yeah, the intersection of people is so strong! Some mentions of exist, reflect.

Good investigations!

cache Didn't read the whole thing, maybe it's good, hard for to take fully rationally with some of their other known stances. It is the smallest page I've downloaded today!

Some cool references. Better scripting language than Python… They point to Roc. I've used Julia before. I'm haskell-curious, though no one mentions that. My top language to learn is Lean, out of curiousity. What's the faster subset version of Python? JAX? That could be good too. All solving problems that I don't really have rn. A full little computer simulator is a cool idea. Could build a better UI for this, like lichess tutorial. Some love for org mode, link to a demo of how to use the bullets.

Gatekeepers of the internet…yes it's annoying when using a VPN for me.

TIL they decrypt the traffic to my domain for the proxy, not technically suprising and it's static site.

• HN

Seems important! Hard to know what is just noise.

Some places to donate to actually do something: ACLU, HRC, Democracy Forward. Oh, internet archive!

• [X] ACLU
• [X] HRC (via Benevity)
• [X] Democracy Forward
• [X] EFF
• [X] Internet Archive

Where to get news? Weekly seems like a decent cadence. Economist? NYT, WaPo seem pretty biased, IMO. Could get the weekly print Economist for $30/month.

Harvard initiative to copy data from all government websites. Sounds complete! Good to have multiple people doing this.

An overview of DeepSeek's technical report. Goes over reinforcement learning and how they got from a base model to the more complex models.

• HN

Idk…

Cool, runs in a VM.

Very cool! Basically a dashboard, and it happens to show Whoop data.

I also learned in the comments here that Will Ahmed is a big Trump supporter. Hmmm.

Lee Robinson: https://leerob.com/n/personal-software.

You can build your own little apps with AI now!

https://www.biotracker.me

Makes a nice-looking dashboard of lab values. Uses LLM processing on a PDF in the background.

People are wary to upload their data to this.

Literally a death clock

> A clock to remind you that life is short. ‘ShortLife’ is a small device showing how much percent of your life is completed based on your life expectancy.

On reddit: How I Built an Open Source AI Tool to Find My Autoimmune Disease (After $100k and 30+ Hospital Visits) - Now Available for Anyone to Use Open source on github: https://github.com/OpenHealthForAll/open-health HN discussion: https://news.ycombinator.com/item?id=42999632

Very cool. Autoimmune seems so hard to diagnose. Other projects like Fasten Health's on prem medical data aggregator are pretty amazing: https://github.com/fastenhealth/fasten-onprem

I could self-host something like that. Though for data format, I'm more inspired by somehow keeping it all in plain text like org mode.

https://ewintr.nl/posts/2025/building-a-personal-private-ai-computer-on-a-budget/?utm_source=tldrnewsletter

• 2 x Nvidia Tesla P40
• HP Z440 Workstation

Pretty cool, way cheaper than buying a macbook with a ton of memory.

https://creston.blog/wasm-will-replace-containers/

Importantly, Cloudflare Workers all run in V8 sandboxes. There are no containers. You have the option of writing your worker in JavaScript/TypeScript, or compiled WASM. A container cannot call another container in the same process. But V8 can. In other words: by deploying WASM in V8 sandboxes, you get all of the developer benefits of microservices with all of the runtime benefits of monoliths. Cloudflare is not the only provider doing this. Wasmer is trying to build a solution in this space as well.

Golden path for software engineering / devops within a company: https://charity.wtf/2018/12/02/software-sprawl-the-golden-path-and-scaling-teams-with-agency/

Good concept! Maybe we can champion a true path at MassMutual. Our LW backend is pretty much on it's own in terms of architecture. Other core stuff all runs on mainframes!

https://pretty-radio-b75.notion.site/DeepScaleR-Surpassing-O1-Preview-with-a-1-5B-Model-by-Scaling-RL-19681902c1468005bed8ca303013a4e2

RL magic is in the air! We introduce `DeepScaleR-1.5B-Preview`, a language model finetuned from `Deepseek-R1-Distilled-Qwen-1.5B` using simple reinforcement learning (RL). It achieves an impressive 43.1% Pass@1 accuracy on AIME2024 (+**14.3% improvement** over the base model), surpassing the performance of OpenAI’s `o1-preview` with just 1.5B parameters. We open sourced our dataset, code and training logs for everyone to progress on scaling intelligence with RL.

• 🌐 [Website](https://agentica-project.com/)
• 👨‍💻 [Github](https://github.com/agentica-project/deepscaler)
• 🤗 [HF Model](https://huggingface.co/agentica-org/DeepScaleR-1.5B-Preview)
• 🤗 [HF Dataset](https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset)
• 📈 [Wandb Logs](https://wandb.ai/mluo/deepscaler-1.5b)
• 🔎 [Eval Logs](https://drive.google.com/file/d/1V_rYKoL35WmubbmWN6PeFg4zo5QOug8X/view?usp=sharing)

Super fit on this particular math task. I am curious what the actual model update step looks like inside of this RL. Seems small enough to really be approachable.

https://rubenerd.com/boring-tech-is-mature-not-old/

Big fan of boring tech, I think there is a whole website and talk about it. Yes: https://boringtechnology.club/

Article says that boring is not just old, or common, but well understood. I think this is just what "boring" actually means! It's a compliment.

Emacs is mentioned on HN.

https://bessstillman.substack.com/p/the-year-i-didnt-survive

Pretty moving story. Her husband Jake died and then she gave birth, not the same person anymore.

https://www.backblaze.com/blog/backblaze-drive-stats-for-2024/

https://news.ycombinator.com/item?id=43013431

They're onto 24TB drives.

I use them to back up! See Backblaze.

Great writing.

https://haterade.substack.com/p/i-tasted-hondas-spicy-rodent-repelling

https://news.ycombinator.com/item?id=43013615

https://desunit.com/blog/marshmallow-test-and-parenting/

Some good notes - just set the example.

Follow through on promises (threats too).

Build a predictable environment.

https://www.worksinprogress.news/p/the-prophet-of-parking

Great article. Parking and streets themsevles are a tragedy of the commons. Maybe it can get better, but I don't think we'll ever really get back to streets as a place for playing!

Incredibly detailed notes on building a retro gameboy basically from scratch. Circuit diagrams and everything.

https://eli.lipsitz.net/posts/introducing-gamebub/

https://www.techtalksweekly.io/p/100-most-watched-software-engineering

Maybe I'll watch some?

https://news.ycombinator.com/item?id=43026590

A couple fun ones:

https://life-stats-eight.vercel.app/

https://shame-roulette.vercel.app/

An idea that has occured to me… At least someone is trying it: https://www.reddit.com/r/QuantifiedSelf/comments/1inq4qy/im_starting_to_think_that_home_assistant_might_be/

New web framework all in JS that makes some interesting choices around where to "hydrate" react pieces. The JS ecosystem still feels pretty crazy.

https://github.com/fullsoak/fullsoak

Really interesting things going on here with rust and linux. Maintainers of core linux packages get to decide which code to allow in and they're generally not allowing Rust. This guy and his team used rust to write a version of linux that runs on arm macs, which is cool, but they can't get their code into linux itself. Also he's taking feature requests quite personally.

https://tesserato.web.app/posts/2025-02-12-CodeWeaver-launch/index.html

I think this is definitely the right idea if you want to have the tools make any useful suggestions.

I think Aider would have to solve this problem…

https://news.ycombinator.com/item?id=43048027

https://aresluna.org/the-hardest-working-font-in-manhattan/

Cool deep dive into typography. The manual cutting machine is awesome.

HN: https://news.ycombinator.com/item?id=43053419

https://vale.rocks/posts/ai-is-stifling-tech-adoption

I agree! But in the same way the existing rich-get-richer mechanisms work.

HN: https://news.ycombinator.com/item?id=43047792

HN: https://news.ycombinator.com/item?id=43053844

Fly.io isn't offer more GPUs. I didn't realize you could rent an A100 for $2.50/hour there. The setup with Fly is so simple… Before I run my own LLM hardware that seems so much simpler to just deploy my own on-demand service to fly. A $5k macbook or $2k linux box are a lot of hours with a $8k GPU!

Self hosted media tracking app. Cool!

Great story about turning the passion project into a legit small company. Good work dude!

https://www.reddit.com/r/QuantifiedSelf/comments/1ipt10f/journaling_semantic_analysis_a_new_angle_on_my/

https://cipher.sysapp.dev/

Just tried to signup for the beta and it didn't work?

https://caddyserver.com/

HN: https://news.ycombinator.com/item?id=43070025

Like nginx, but the auto-https would be nice. Can even use it to better handle local traffic with HTTPS turned on.

People also mention traefik. Integrates nicely with tailscale: https://tailscale.com/blog/caddy

https://uchu.style/

They look nice!

Oklch color space: https://developer.mozilla.org/en-US/docs/Web/CSS/color_value/oklch

• https://tailwindcss.com/docs/colors
• https://huetone.ardov.me/
• https://github.com/darekkay/a11y-contrast
• https://darekkay.com/blog/accessible-color-palette/

Developer, NetOperator Wibby, has an interesting website: https://webb.page/

http://hforsten.com/homemade-polarimetric-synthetic-aperture-radar-drone.html

I'm always amazed by people that can do hardware.

For nate:

https://www.ebay.com/sch/i.html?_nkw=fiestaware+dinner+plates+radioactive&_sacat=0&_from=R40&_trksid=p2334524.m570.l1313&_odkw=fiestaware+dinner+plates&_osacat=0

I think we have one around the house somewhere…how does stuff just disappear?

HN: https://news.ycombinator.com/item?id=43073969

Source: https://kindlemodding.org/jailbreaking/WinterBreak/

https://weeks.ginatrapani.org/

https://news.ycombinator.com/item?id=43061498

I must have covered this before - she points to the original references.

Study: https://e360.yale.edu/digest/brake-pads-lung-damage-study

Also tires are bad, which will be worse with electric cars.

I think most people would be shocked to learn it's not just the exhaust…

HN: https://news.ycombinator.com/item?id=43058993

Stories of people next to roads getting stuff covered in black dust.

HN: https://news.ycombinator.com/item?id=43059579

It's possible! But probably better to just buy a graphics card.

The self hosting seems fun but I'm good just using Claude.

Benchmarks for different hardware: https://github.com/geerlingguy/ollama-benchmark?tab=readme-ov-file#deepseek

Made the front page of HN and I've been meaning to try it.

HN: https://news.ycombinator.com/item?id=43063167

I set up synology's media server to try that…

Our vizio TV's only option is plex though!

Great post from Steven Johnson. Covers

• Herbert Simon's lecture on information abundance
• Chris Hayes' book The Siren Call
• A biography writer's process
• SNL skit parodying NotebookLM

I really like this quote from Simon, that Hayes covered, which summarizes the core idea:

What makes a given information processing system useful to an organization isn’t how much information it generates or even the raw amount of information it can process. Rather, [Simon argues], “the crucial question is how much information it will allow to be withheld from the attention of other parts of the system… To be an attention conserver for an organization, an information-processing system must be an information condenser.”

https://pubmed.ncbi.nlm.nih.gov/26607247/

20-30 degrees is ideal for energy consumption. No suprises here, having considered the math and done some personal VAM tests thinking about everesting. That's degrees, so it's really 40-60% grades by percentage.

Love these bikes! Firefly is represented.

https://lovelybike.blogspot.com/2012/08/the-bikes-of-d2r2.html

https://www.petecodes.io/marketing-genius-of-bryan-johnson/

He's good at marketing!

Interesting startup doing health assessment: Neko Health. Some sort of full body scan.

https://spyglass.org/consumer-ai-services-apps/

Comparison of consumer AI models.

Still have OpenAI on the top, which I guess it really is. A lot of pretty similar models nowadays.

Startup by OpenAI CTO https://en.wikipedia.org/wiki/Mira_Murati.

https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/

Mix of all of the cool tools: aider, cursor, etc. Hadn't heard of remix.

HN

Really good for greenfield projects, for sure.

Recommended: "Co-Intelligence: Living and Working with AI" By Ethan Mollick

https://www.bitecode.dev/p/a-year-of-uv-pros-cons-and-should

A TODO.

HN

You can pin versions by time, very cool.

#!/usr/bin/env -S uv run
# /// script
# dependencies = ['requests', 'beautifulsoup4']
# [tool.uv]
# exclude-newer = "2023-10-16T00:00:00Z"
# ///
import requests
from bs4 import BeautifulSoup

rmlint will replace duplicate files with reflinks (if your fs supports them — xfs and btrfs do), or hardlinks if not.

Cool IDE with built in viz for Python. But it's $20/month! Need a free tier, IMO.

HN

I was running my tailscale DNS on it for a long time. The only thing I really noticed is that it blocked google ad links, which I often wanted to click.

HN discussion

https://jazco.dev/2025/02/19/imperfection/

Good engineering story. Solving data problems at scale, pretty cool.

But as user, I would want all of the posts on my timeline? I wouldn't want to miss any. I guess that if there are 1000 posts per day, I'm not reading all of them anyway.

HN

https://news.ycombinator.com/item?id=43103073

Interesting… But I'm not building an AI powered front end app anytime soon.

https://github.com/mastra-ai/mastra

https://www.scientificamerican.com/article/broken-legs-and-ankles-heal-better-if-you-walk-on-them-within-weeks/

Lots more anecdata on HN:

let pain be your guide

On AI-based weapons systems

Reddit QS.

Does anyone know of a unified health api for Apple Healthkit, Google Fit, Withings, etc? Or at least as many as possible. I know there are paid ones, but does someone know one that is free or open source?

Or something similar to Health Auto Export but for more data sources.

Terra, fitnesssyncer

Really really great guide on how to get sleep working.

https://www.pokemonsleep.net/en/devices/

https://openeurollm.eu/

Just getting started, nothing done. There's also EuroLLM, which does have some models.

HN

Will be interesting once it's actually released?

https://news.ycombinator.com/item?id=43124018

https://www.figure.ai/news/helix

Maybe LLMs will help them generalize a bit. Who knows…

HN

Yeah you have to trust models like source code…

https://blog.sshh.io/p/how-to-backdoor-large-language-models

HN

I don't see how this won't backfire. It's not the big budget items…

Yet:

Very cool 3d explorer of the radio tower in san fran.

https://www.trainerroad.com/forum/t/jonathan-on-the-nero-show/100682

Interesting podcast.

https://www.pnas.org/doi/10.1073/pnas.2420498122

Like a neck gaiter with strain gauges to see deflection. Pretty cool. Would be dirt cheap to make.

https://www.nature.com/articles/s41592-024-02582-2

Projecting onto a manifold, which is pretty cool. Definitely abstractions/mental models are key.

https://trufflesecurity.com/blog/removing-jeff-bezos-from-my-bed

Clickbait title, it's eightsleep. They use AWS to stream data.

Some bad engineering decisions, but they couldn't really ssh into his house wihout exposed NAT!

The aquarium chiller is nice.

HN: https://news.ycombinator.com/item?id=43129439

Some good HA stuff in there on HN. Esphome, etc. Zigbee hub: https://eu.aqara.com/products/aqara-2k-security-indoor-camera-hub-g3

The smart cube for screen free control is very cool idea.

https://berthub.eu/articles/posts/you-can-no-longer-base-your-government-and-society-on-us-clouds/

Yeah, EU doesn't feel like AWS won't provide their data to Elon. Was never a good idea.

FFMpeg gets legit speedups from assembly and SIMD.

https://github.com/FFmpeg/asm-lessons/blob/main/lesson_01/index.md

https://www.openbsd.org/innovations.html

HN: https://news.ycombinator.com/item?id=43143777

Cool. Need to finish my format and get listed up here!

https://indieblog.page/

HN: https://news.ycombinator.com/item?id=43139953

https://www.mooreds.com/wordpress/archives/3518

Make the answer you want to get the easier thing to do.

HN: https://news.ycombinator.com/item?id=43144611

You can do this without being aggresive…

Backlinks: Johnny Decimal

https://johnnydecimal.com/10-19-concepts/12-advanced/12.03-the-standard-zeros/

HN: https://news.ycombinator.com/item?id=43128093

Lots of people have the own tools/workflow, of course

• Personal inventory app: https://github.com/mo42/inven
• Rewind AI has a wearable mic: https://www.limitless.ai/#pendant
• Don't organize at all: https://news.ycombinator.com/item?id=43135927
• PARA tool https://thoughtscape.app/ (https://news.ycombinator.com/item?id=43136082)
• NetOpWibby's JD script: https://gist.github.com/NetOpWibby/7e39068c1d0209e4412e3a05e800120c
• Karl Voit's: https://karl-voit.at/folder-hierarchy/
• Hamster system: https://github.com/slowernews/hamster-system

Pretty cool story. Dedication!

https://news.ycombinator.com/item?id=43133174

https://freakathlete.co/products/cable-attachment

I haven't hooked up the leg attachment yet, don't need this.

It didn't change behavior much.

Link: https://www.nature.com/articles/s41562-024-02032-1

https://github.com/superglue-ai/superglue

Very cool idea. Use LLMs to validate data mappings.

Good on you, EA.

https://news.ycombinator.com/item?id=43197131

Interesting project. Andreas Kling.

HN post about it, promising.

https://nroottag.github.io/

You can piggy back on Apples network. They somehow patched this for new devices.

Maybe open airtags: https://news.ycombinator.com/item?id=42835772

https://github.com/semaphoreio/semaphore

https://news.ycombinator.com/item?id=43186406

Pretty awesome.

https://arstechnica.com/gadgets/2025/02/framework-known-for-upgradable-laptops-intros-not-particularly-upgradable-desktop/

Most desktops are more swappable, so this is a departure from their laptops.

HN points out that the soldered ram is for better performance, but still?

https://hypercritical.co/2025/02/25/hyperspace

Interesting app that does compression on your hard drive effectively by hard linking parts of files that are the same. Cross-file compression. Pass for me.

HN: https://news.ycombinator.com/item?id=43173462

https://sockpuppet.org/blog/2025/02/09/fixing-illinois-foia/

Very fun story. Schemas are not sensitive information: they're needed to make the right queries.

HN: https://news.ycombinator.com/item?id=43175628 Matt Chapman is on HN!

Olson and I both need them.

https://www.schwalbetires.com/Marathon-Winter-Plus-11100597.01

./static/2025-02-28-iq-tests.html

Really great research on IQ, the best I've seen. It's definitely measuring something.

From Every.to. A potentially really good use of AI/ML: scrape N profiles, and then use the LLM to generate content from them.

https://www.ericdaigle.ca/posts/breaking-into-dozens-of-apartments-in-five-minutes/

Yeesh. The software should make you change the default password…

HN: https://news.ycombinator.com/item?id=43160884

https://jiha-kim.github.io/posts/introduction-to-stochastic-calculus/

To read!

HN: https://news.ycombinator.com/item?id=43160779

app actually verifies you touch grass!

https://touchgrass.now/

It must be a joke. Wouldn't work in the snow here!

HN: https://news.ycombinator.com/item?id=43158660

I guess it's real. Lol.

https://www.ifixit.com/News/108371/right-to-repair-laws-have-now-been-introduced-in-all-50-us-states

Good! Particularly for the people doing the repairs (average consumer doesn't repair).

https://news.ycombinator.com/item?id=43161777

https://hvpandya.com/sabbatical

Sounds nice to take a break! Not too deep of thoughts here, but good for them.

https://github.com/anthropics/claude-code

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

Maybe like Aider? I definitely need to try this.

https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview

Okay it's so easy to set up. Have to use the API, feels like I shouldn't pay for both the pro web UI AND API credits.

HN: https://news.ycombinator.com/item?id=43163011

Tried it out on a few PRs. Pretty great honestly.

https://blog.6nok.org/tailscale-is-pretty-useful/

I agree! Used it to connect up my local LLM even.

HN: https://news.ycombinator.com/item?id=43270835

Zero-trust is interesting. With things like OpenZiti and https://zrok.io/. Cloudflare tunnels are more zero-trust.

512GB of memory its pretty great. Could run AI with ease.

On my exact M1 pro 16gb someone is running this well:

LM Studio with the deepseek-r1-distill-llama-8b model @ Q8₀

Also sounds nice, just for carrying it around.

https://google-research.github.io/self-organising-systems/difflogic-ca/?hn

Fancy compression?

Retire is misnomer!

Paper: https://www.science.org/doi/full/10.1126/sciadv.ads1560?af=R

HN: https://news.ycombinator.com/item?id=43279494

https://news.ycombinator.com/item?id=43280615

Cool! I should see if I can speedup the core loop of the points rewards via Cython.

https://blog.startifact.com/posts/succinct/

https://www.natesilver.net/p/a-few-words-about-fivethirtyeight

Nate had left a couple years ago, working on a related project it sounds like.

https://dynomight.net/theanine/

Thoughtful self experiment. Didn't seem to work for him. A few resources for evaluating supplements:

• European food safety: https://www.efsa.europa.eu/en/efsajournal/pub/2238
• Examine: https://examine.com/supplements/theanine/

Came up on HN <2025-03-09 Sun> https://news.ycombinator.com/item?id=43305803

Some praise, some not praise.

• Stress definitions https://pmc.ncbi.nlm.nih.gov/articles/PMC8434839/
• Dust mite guide: https://dustmiteguide.com/ Action is just to keep humidity below 50%, target 40%. I thought our humidity was too low… Hard to see those numbers in winter in new england.

https://geohot.github.io//blog/jekyll/update/2025/03/03/demoralization-is-just-beginning.html

He's got two suggestions:

1. Brain drain the world. Work visas for every person who can produce more than they consume. I’m talking doubling the US population, bringing in all the factory workers, farmers, miners, engineers, literally anyone who produces value. Can we raise the average IQ of America to be higher than China?
2. Back the dollar by gold (not socially constructed crypto), and bring major crackdowns to finance to tie it to real world value. Trading is not a job. Passive income is not a thing. Instead, go produce something real and exchange it for gold.

Maybe? I think gold is long gone.

HN: https://news.ycombinator.com/item?id=43261729

Could be fun to run a mail server?

https://www.xmox.nl/

Could I have this running for a sub domain? (I don't think subdomains support email…. how would the MX records work?)

https://news.ycombinator.com/item?id=43259182

I've been saying for awhile that LLMs are just compression.

Hutter prize: http://prize.hutter1.net/ Is just compression!

Pretty cool!

https://news.ycombinator.com/item?id=43255529

One that you can buy: https://hellonuio.com/

DIY kit: https://shop.beekeeb.com/product/lets-split-v2-keyboard-pcb-kit/?srsltid=AfmBOoqSnweOL2oUq39ST7nfdU0dlVJflhWocliGm2S4aNoTssoexW4y

Firebase

https://news.ycombinator.com/item?id=43210858

https://x.com/EthanEvansVP/status/1895845734177452369

Interesting points here. RTO makes sense if you have a driver and a nanny.

Cool.

https://www.kaseyhou.com/#/repairable-flatpack-toaster/

Time is infinite until it isn't.

https://developer.chrome.com/blog/command-and-commandfor

More functionality in HTML, which is cool. Basic sites can use just HTML without needing JS. I like the examples in the article comparing popular frameworks.

HN: https://news.ycombinator.com/item?id=43292056

HN: https://news.ycombinator.com/item?id=43302073

I haven't tried Kagi yet but heard lots of good things about their search tools. The small web one in particular.

https://usher.dev/posts/2025-03-08-kill-your-feeds/

HN: https://news.ycombinator.com/item?id=43302132

this is basically a summary of the most recent Technology Connections video, Algorithms are breaking how we think: https://youtu.be/QEJpZjg8GuA

• https://indieblog.page/random
• https://wiby.me/surprise

HN: https://news.ycombinator.com/item?id=43307996

Sounds like just a terrible idea to stop providing support like this. Who will keep buying from the US?

https://raoulpop.com/2012/04/15/chasing-rfi-waves-part-seven/

HN: https://news.ycombinator.com/item?id=43315406

The radio astronomy sounds super cool. I couldn't have been so far from here at Virginia Tech.

https://randomaugustine.medium.com/on-apple-exclaves-d683a2c37194

HN: https://news.ycombinator.com/item?id=43314657

This goes way above my head but it's a pretty interesting set of ideas.

https://www.science.org/doi/10.1126/sciadv.adt1646

We review the literature on the gender gap in well-being, identifying two key paradoxes. First, although women today report higher levels of life satisfaction and overall happiness than men, they experience worse outcomes in mental health and negative affect. Second, despite substantial advances in women’s social and economic status over the past 50 years, their well-being relative to men has declined. We explore the evidence supporting these paradoxes, considering potential explanations related to differential expectations, biology, and scale use. Using global data from 2006 to 2023 and long-term data from Europe and the US since the 1970s, we provide empirical illustrations. These findings reveal a diverse and seemingly inconsistent pattern of gender well-being gaps between countries, suggesting that the first paradox is not universally applicable. However, there is clear global evidence of a relative decline in women’s well-being, particularly in terms of negative affect.

Powerful introduction too.

https://www.aliexpress.us/item/3256802887739257.html?aff_fcid=99c2007c63074f1a873616c5815dcabc-1741647027713-00420-_on3jcZo&tt=CPS_NORMAL&aff_fsk=_on3jcZo&aff_platform=shareComponent-detail&sk=_on3jcZo&aff_trace_key=99c2007c63074f1a873616c5815dcabc-1741647027713-00420-_on3jcZo&terminal_id=0a5c17e492d04a37840093c509633668&afSmartRedirect=y&gatewayAdapt=glo2usa4itemAdapt

Since the link will probably disappear, here's the cache.

Also a pair of maybe fake princeton carbonworks version for less than 1k!

Open source book: https://github.com/MathFoundationRL/Book-Mathematical-Foundation-of-Reinforcement-Learning

HN: https://news.ycombinator.com/item?id=43323946

Want to read it someday…

https://exclusivearchitecture.com/03-technical-articles-CLT-12-autofocus-systems.html

Very cool technology…

HN: https://news.ycombinator.com/item?id=43320230

https://www.seven39.com/

Fun idea! Kind of like beReal.

HN: https://news.ycombinator.com/item?id=43328095

https://jackhopkins.github.io/factorio-learning-environment/

HN: https://news.ycombinator.com/item?id=43331582

Interesting idea… RL should be able to crush this particular task thought.

https://github.com/ZachGoldberg/Startup-CTO-Handbook/blob/main/StartupCTOHandbook.md

HN: https://news.ycombinator.com/item?id=43337703

Seems like there's going to be some good advice in here.

From the DIY Cold Plunge guy.

Sorry Joe, it looks like a coffin!

https://diycoldplunge.com/products/diy-infrared-sauna-plan

https://github.com/johnbean393/Sidekick

OSX app to be a LLM-based research assistant / knowledge base. Cool idea. Don't love this implementation.

New ones are out.

https://store.repebble.com/

No GPS though - which is definitely the feature that I want.

A $3k home computer that can easily run AI models! Pretty cool. But I'd have to spend a hell of a lot on cloud resources to make this worth it.

https://www.scotthyoung.com/blog/podcast/

I loved reading his stuff when I was in college. Very inspiring.

Glad he's still going strong. He has a kid!

HN: https://news.ycombinator.com/item?id=43398518

It's a cloud security scanning company. A buy for the customers, maybe more than the tech.

Yes but this is also half the point. It should be easy and secure for users, this doesn't come tradeoff-free.

https://ericmigi.com/blog/apple-restricts-pebble-from-being-awesome-with-iphones

On the other hand, I'd love to be able to reply from my garmin watch.

HN: https://news.ycombinator.com/item?id=43401245

HN: https://news.ycombinator.com/item?id=43390400

https://arxiv.org/abs/2503.02113

To read.

Julia Zimmerman's thesis.

Pretty amazing slides, love the artwork: https://docs.google.com/presentation/d/1xW2JMfDFGGF9Rszra6nU4sLjaMx-pfqMNoMPkJfDKrs/edit#slide=id.p

Cache: ./static/zimmer2025language.pdf

https://www.astralcodexten.com/p/the-ozempocalypse-is-nigh

The generics are going away. Bummer for people.

Making money on pharmaceuticals that are expensive to develop but then saves lives seems like an ethical problem from the start.

https://www.benkuhn.net/pjm/

Some good advice in here. It's hard work, requires focus.

https://techcrunch.com/2025/03/14/tern-ais-low-cost-gps-alternative-actually-works/

Accelerometer based? Cool idea?

Sounds like it uses accelerometers and road network data to update it's position. Align left/right turns to the road's turns to help with dead-reckoning.

Would love to try this for cycling.

Backlinks: Docs

HN: https://news.ycombinator.com/item?id=43378239

https://github.com/suitenumerique/docs

I'm a sucker for knowledge management solutions. On man, it uses Django!

Competitor to outline.

Will add a backlink to my list.

Cool idea.

https://news.ycombinator.com/item?id=43342361

There are long lists of these things:

https://gist.github.com/SMUsamaShah/fd6e275e44009b72f64d0570256bb3b2

Even more on the HN.

https://github.com/kern/filepizza

https://news.ycombinator.com/item?id=43334190

Need to check this out…

https://medium.com/@caitlin9165/fastplotlib-driving-scientific-discovery-through-data-visualization-418f8bff094c

https://github.com/vispy/jupyter_rfb

Looks like fastplotlib would be nice for doing 3d graphics that can leverage the GPU.

Love this. HN: https://news.ycombinator.com/item?id=43342712

https://lwn.net/Articles/1015402/

Pretty cool - able to build the same binary from the same source code. Controlling for all of the OS differences and randomness.

Here's the actual update: https://lists.reproducible-builds.org/pipermail/rb-general/2025-March/003675.html

HN: https://news.ycombinator.com/item?id=43484520

VanillaOS builds on Debian: https://vanillaos.org/

HN: https://news.ycombinator.com/item?id=43485649

People don't feel safe at the borders! Among other concerns (tariffs, threats of annexation).

Very cool achievement.

HN: https://news.ycombinator.com/item?id=43483660

Details on their satellite: https://spaceinafrica.com/2025/03/15/botswana-successfully-launches-first-satellite-botsat-1/

https://github.com/medialab/xan/blob/master/docs/LOVE_LETTER.md

A decent attempt to defend CSV! Having dealt with so much pain from poorly formatted CSVs I can never look back from parquet.

I didn't realize the double quote trick to escape quotes. Clever!

HN: https://news.ycombinator.com/item?id=43484382

> I so hate CSV.

Yeah. I'm so with you dude.

https://cardinalnews.org/2025/03/28/i-drove-300-miles-in-rural-virginia-then-asked-police-to-send-me-their-public-surveillance-footage-of-my-car-heres-what-i-learned/

In virginia. I know those town names! "Flock" camera data is subject for FOIA which seems bad. Though, better than not being subject to it. Best to not have the data exist in the first place.

HN: https://news.ycombinator.com/item?id=43504413

Flock was a YC17 company!

You could do it yourself: https://github.com/jfh01/diy-alpr Questionable legally to just post this for free online.

HN: https://news.ycombinator.com/item?id=43517833

May not be such a bad idea, actually. Just brush your teeth.

HN: https://news.ycombinator.com/item?id=43517301

They used Jump Studio to get the connection.

https://pagedout.institute/download/PagedOut_006.pdf#page=3

Pretty amazing magazine. The rust in python example is great. On a single page of PDF a full working example…page 49: https://pagedout.institute/download/PagedOut_006.pdf#page=49

They're taking submissions. Had to fit in a single page.

https://www.dcrainmaker.com/2025/03/komoot-acquired-history-says-this-wont-end-well.html

I actually hadn't heard of them. Got my first komoot links just yesterday, VTXL 4-day in June.

Backlinks: Cycling

TR forum: https://www.trainerroad.com/forum/t/cn-40mm-road-tyres-are-faster-for-nearly-everyone/101891

GP5000's are good road tires. Some hate for the specialized roubaix, which I really liked, and love for the schwalbe pro. Schwalbe pro road tires: https://www.schwalbetires.com/Pro-One-11653978.

Would love to build out the flock of bikes more completely, having road/tt/gravel/mtb all ready to roll.

• New rear road bike tire (current is shot) - ordered with cassettes and chains.
• New rims for these wheels? Maybe?
• Dedicated bike for trainer (it's not compatible with zwift ride). Put on the tarmac, works.
• Better wheelset for TT?
• Casettes/chains shot on tarmac and TT bikes - ordered to laughing dog, will come 5/14.
• MTB back as a MTB.
• Fenders for FF1169 - talked to FF about this.

Made some moves on this list.

Backlinks: Personal data infrastructure

https://karpathy.bearblog.dev/digital-hygiene/

This VERY closely resembles my own. Really nicely written up.

Backlinks: Mechanical Keyboards

https://www.norbauer.co/products/the-seneca

If you have more money than you know what to do with…

https://refactoringenglish.com/tools/hn-popularity/

https://luj.fr/blog/how-nixos-could-have-detected-xz.html

Bold claim. I still want to learn this Nix thing.

HN: https://news.ycombinator.com/item?id=43448075

Fun: https://lexfriedman.com/friedman

https://busy.bar/

HN: https://news.ycombinator.com/item?id=43643534

From the folks at flipper. Total overkill at $250 and with the features.

Can see how it would be helpful for open concept like Springfield office.

Seems like a useful resource.

https://simonmonk.org/tyee7

HN: https://news.ycombinator.com/item?id=43589776

Maybe I could build something like https://dmitry.gr/?r=05.Projects&proj=36.%208pinLinux

Love little hacker projects like this

https://pico.sh/

HN: https://news.ycombinator.com/item?id=43560899

https://emacsredux.com/

I can always get better at Emacs.

HN: https://news.ycombinator.com/item?id=43546004

I should really give a fully functional language a go. Might change how I see things.

HN: https://news.ycombinator.com/item?id=43625474

https://yuxi-liu-wired.github.io/essays/posts/cyc/

To read.

https://nothinghuman.substack.com/p/the-tyranny-of-the-marginal-user

cache

Enshittification, one mechanism.

https://protesilaos.com/

Very interesting blog!

Including an emacs book. He maintains about 100 emacs packages.

https://www.gleech.org/llms

Interesting blog. Very gwern-esque.

https://blog.jim-nielsen.com/2025/love-for-interoperability/

Nice post: having plain text or even a standard (email) means you can move data around.

Yes!

File over app

https://vitalik.eth.limo/general/2025/04/14/privacy.html

privacy is less needed for people whose life situations are relatively normal, and more needed for people whose life situations deviate from the norm, in any direction. And once you add up all of the different directions that matter, the number of people who really need privacy ends up being quite a lot - and you never know when you will become one of them. This is a big reason why privacy is often underrated: it's not just about your situation and your information today, it's also about the unknown unknowns of what happens to that information (and to how it affects you) going forward forever into the future.

A lot of really great points here. Democracy depends on private votes.

https://www.geoffreylitt.com/2025/04/12/how-i-made-a-useful-ai-assistant-with-one-sqlite-table-and-a-handful-of-cron-jobs

HN: https://news.ycombinator.com/item?id=43681287

Pretty great!

Here's the code on val town: https://www.val.town/x/geoffreylitt/stevensDemo/code/populateDemo.ts

https://github.com/arpanghosh8453/garmin-grafana

Uses the garmin auth creds to scrape out the data.

Pure andy-bait.

https://emilygorcenski.com/post/owning-my-own-data-part-1-integrating-a-self-hosted-calendar-solution/

HN: https://news.ycombinator.com/item?id=43643343

Add a link with C-c C-l and autocomplete on "attachement:" and you get:

2025-04-16-Opinion ｜ How to Be a Happy 85-Year-Old (Like Me) - The New York Times (4₁₅₂₀₂₅ 10：15：48 PM).html

Org mode is not on his list, but maybe it should be!

Love this. They were good! Of course. Very human story.

Good designs expose systematic structure; they lean on their users’ ability to understand this structure and apply it to new situations. We were born for this.

Bad designs paper over the structure with superficial labels that hide the underlying system, inhibiting their users’ ability to actually build a clear model in their heads.

Very great point. Should try reading the book.

Interesting article.

Not going to take over the world any time soon.

AI benchmark to manage a vending machine business.

Some of the model succeed, sometimes! Need to be able to keep track of context in a more coherent way.

https://github.com/humanlayer/12-factor-agents

I'll revist this when we do the agentic coding week.

https://blog.atuin.sh/atuin-desktop-runbooks-that-run/

Atuin syncs shell history across devices. I can see how this is valuable!

The desktop is a notebook environment, looks like jupyter. But they're at the shell. Tech:

We're using Tauri (https://v2.tauri.app/) on the client, and Elixir + Phoenix (with a little bit of Rust via Rustler) on the server

Basically yeah jupyter notebooks: https://www.nurtch.com

This uses markdown as the backend: https://runme.dev/

Another jupyter competitor: https://github.com/marimo-team/marimo

HN suggests Org babel, filed this under org mode.

Pretty neat. Hardware hacking is always fascinating.

https://github.com/PiLiDAR/PiLiDAR

HN: https://news.ycombinator.com/item?id=43738561

Interesting…

https://www.trainerroad.com/forum/t/strava-acquires-runna/102615

Feeling anti-strava these days, with the API lockdown and such.

They are the WORST.

Ugh: https://prospect.org/power/2025-04-17-intuit-turbotax-wins-battle-against-taxpayers-irs-direct-file/

Glad I gave them $0 this year, instead of the $270 they wanted.

Backlinks: Long term outlook

Read this through. I agree with most of their points. In particular - I don't see any evidence of AI bootstrapping AI research, it's not even that helpful in my own day to day (which is less complex than AI research).

There exists an AI incident database: https://incidentdatabase.ai/

On the limits of predictability: Measuring the predictability of life outcomes with a scientific mass collaboration.

https://new.computer/

Connected to your calendar, can become a pretty powerful tool. Found this while researching Stevens.

Playing with it now.

Components make pretty powerful framework-free websites: https://plainvanillaweb.com/index.html

HN: https://news.ycombinator.com/item?id=43954896

Framework free reasonably complex site: https://restofworld.org/ Oh gosh, well it uses wordpress. Lol.

HN: https://news.ycombinator.com/item?id=43925732

• Of course they mention mcmaster: https://www.mcmaster.com/
• RockAuto is a good one, that I've used.
• Small parts: https://shop.sdp-si.com/

Nice:

I call it Trader UI vs VC UI

A couple HN alternatives:

• https://hcker.news/
• https://hckrnews.com/

I also think this is what FitIQ is trying to provide.

Guy's resume/info as an MCP server: https://ai.jakegaylor.com/

Kind of the like the "me API": https://www.bobapi.com/

https://raz.sh/blog/2025-05-02_a_critical_look_at_mcp

My conclusion is that the whole suggested setup for HTTP transport (SSE+HTTP and Streamable HTTP) should be thrown out and replaced with something that mimics stdio… Websockets.

Stdio is a lot like websockets…they seem fine!

https://news.ycombinator.com/item?id=43906841

Neat idea…kind of like google brain or semantic search database.

https://blog.rahix.de/design-for-3d-printing/

HN: https://news.ycombinator.com/item?id=43888117

Good resource on 3d printing. Some tricks that you wouldn't guess.

They dropped it to 5 exercises:

https://www.army.mil/e2/downloads/rv7/aft/AFT_Scoring_Scales_250601.pdf

My max/min for 100/60 points on each are (and to "pass", need 350 total, not just the min on each):

• MDL 350/140
• HRP 60/13
• SDC 1:33/2:36
• PLK 3:25/1:15
• 2MR 13:42/20:44

MDL for 3 reps, hex bar. SDC is 25 meters with 5 legs of "down and back":

• Sprint
• Drag 90lb sled
• Side shuffle (legs can't touch/cross)
• 40lb weight each hand
• Sprint

Sounds qute fast to do this in 90 seconds, it's 250 meters!

Nice write up of the history and such. Basically a markdown editor and a database, which is cool.

https://amberwilliams.io/blogs/building-my-own-pkms

HN: https://news.ycombinator.com/item?id=44022448

Dataview in Obsidian is a lot like the agenda of org mode: https://obsidian.rocks/dataview-in-obsidian-a-beginners-guide/

Okay I love this.

npx ashleywillis

https://ashley.dev/posts/turning-feedback-into-features/