jasonthorsness

While maybe the model could do everything from first principles every time, once you have a known good tool that performs a single action perfectly, why not use that tool for that action? Maybe as part of training, the model could write, test, and learn to trust its own set of tools, rather than rely on humans to set them up afterwards.

I used Gemini CLI and a few rounds of Claude CLI at the beginning with stock VSCode

I used Gemini instead of my usual Claude for a non-trivial front-end project [1] and it really just hit it out of the park especially after the update last week, no trouble just directly emitting around 95% of the application. Now Claude is back! The pace of releases and competition seems to be heating up more lately, and there is absolutely no switching cost. It's going to be interesting to see if and how the frontier model vendors create a moat or if the coding CLIs/models will forever remain a commodity.

[1] https://github.com/jasonthorsness/tree-dangler

The community can make whatever UI it wants, using the API:

https://github.com/HackerNews/API

I wrote one for myself:

https://hn.unlurker.com

Or there are various ways to restyle the main site with a browser script.

Had to read the article to find how the mice were lifting!

"To access food, mice were fit with a small collar around the shoulders that required them to lift and set down the cage lid in a squat-like motion, which activates concentric and eccentric muscle contraction"

There is a diagram as well :)

AES-GCM sensitivity to nonce reuse is a tricky implementation detail. Here they acknowledge it but then don’t share their solution - and in fact the header contains 16 bytes for the nonce instead of the expected 12 bytes and they do not share what bytes are random. Did I miss something, anyone know?

"Starting today, GPT‑5.1-Codex-Max will replace GPT‑5.1-Codex as the default model in Codex surfaces."

Wow, I spent last weekend using a tag-team of Claude and Codex and found Codex to more often get better results (TypeScript physics/graphics application). I probably only wrote a few hundred lines of code out of many thousands; it did a really good job.

Now I guess I'll ask the new Codex to review the work of the old!

I tried this as well at https://github.com/jasonthorsness/ginprov (hosted at https://ginprov.com). After a while it sort of starts to all look the same though.

Both of my kids wanted to need a glasses prescription

It was early. Social acceptance will happen like it did for phones (although I still get weirded out looking at a bus full of people all staring at a device).

When I started programming debuggers were very good. Windbg for example was incredibly powerful, you could debug the Windows kernel, boot process, run scripts, author plugins, anything.

And yet here decades later this list about debugging doesn't even mention a debugger, and in many environments they are worse and harder to use than what we had before. I'm so disappointed!

I don't wear glasses but that's the primary form factor I can imagine. Already socially and legally acceptable, non-audible communication channel plus low-volume audio, points in a direction that generally matches attention. And pre-AI investment in augmented reality has given it a head start.

Yeah the encouragement of shadowing is a little weird (learning Rust coming from Go where it is sort of discouraged)

It’s great that these reproducible builds are possible and this is an incredibly thorough and careful validation. Thanks!

Lots of articles about how solar panels are even less expensive than traditional fence panels; it's incredible. If I had a suitable property I would be installing them even in Seattle area where it's cloudy and dark for a significant part of the year (summers are pure sun though!).

Thanks for the book recommendation! Heard about it a few times now so I'll check it out.

Money alone cannot win the race; it will require a strong vision, courage, and luck. OpenAI and XAI and Anthropic (though underfunded) seem to lead in those factors so far.

NaNs aren't always equal to each other in their bit representation either, most of the bits are kept as a "payload" which is not defined in the spec it can be anything. I believe the payload is actually used in V8 to encode more information in NaNs (NaN-boxing).

The 6-hour counter is most certainly, painfully true. If you work with an AWS rep please complain about this in every session; maybe if we all do they will reduce the counter :P.

Some reasons for this:

1. Free software is developed for the developer's own needs and developers are going to be power users.

2. The cost to expose options is low so from the developer's perspective it's low effort to add high value (perceiving the options as valuable).

3. The developer doesn't know who the customer is and rather than research/refine just tries to hit all the boxes.

4. The distribution of the software itself means anyone who successfully installs it themselves really is a power user and does like the options. Installing it for family and friends doesn't work.

Probably many other factors!

Ahh it got me, Alaska air web site has an Azure outage banner

Using a single DBMS for many purposes because it is so flexible and “already there” from an operations perspective is something I’ve seen over and over again. It usually goes wrong eventually with one workload/use screwing up others but maybe that’s fine and a normal part of scaling?

I think a bigger issue is the DBMS themselves getting feature after feature and becoming bloated and unfocused. Add the thing to Postgres because it is convenient! At least Postgres has a decent plugin approach. But I think more use cases might be served by standalone products than by add-ons.

the entire page has horizontal lines washed over it

One thing I have observed (but this is in B2B) is that developers can be a lot more cautious around asking for money than they should be. Sales can often tell customers they need to pay and they will pay. Complaints about cost are part of the negotiation process in this environment. In B2C I don't know.

This attitude doesn't reflect poorly on the users. Outside any kind of charity situation, if you pay for something that is free it makes you a sucker. You need to make sure users understand that paying makes them savvy/smart or otherwise benefits them.

Apparently some sizes are cursed!

It would be great to repeat the author’s tests on other CPU models

EVs often have great acceleration, so I would think it would be a natural fit for the sports car market. Even the ordinary ones are a lot more fun to accelerate on a freeway onramp than an ordinary 4- or 6-cylinder gas engine.

Article makes a huge deal about humming/buzzing but there’s no way that’s a real thing right? What would be the reason it would go beyond like any HVAC-ish thing on a large building? Lost me at the first sentence :(

I’ve had a heat pump with backup gas in Seattle area since 2016. Great to have AC and heat; the gas comes on when house needs to be warmed quickly. I like constant air circulation and have it set so fan runs even when it doesn’t need to heat or cool so it works really well for me; no complaints at all.

I’ve been learning Rust via the book and a great article I found on linked lists [1]. Coming from C++ the lifetimes/borrows concepts make sense at a high level but the practical details seem to get pretty crazy. If anyone here knows Rust well does the OP article have a good take or is it missing something?

[1] https://rust-unofficial.github.io/too-many-lists/