Gil Noh (노태길)

@simonw I feel like we need to stack these tests now

@ZPostFacto Doesn't need to be toddlers. It's enough to know that some people will press blue to justify pressing blue to save them. That's the only option that saves everyone, not this "everyone just press red" fantasy selfish cunts use to rationalise throwing everyone else under the bus.

I really love Claude, and convinced our company to migrate to Antrophic, and generally happy about it... but, recent events make me feel like, we cannot trust Antrophic as reliable provider - e.g. never make long commitment, always look for alternatives.

Secret of the successful people 😃 -- They try (sample) without replacement! Darn I wish I could!!

@bcherny @GergelyOrosz This easily reproducible thing was not even tried, and the reaction was to point "the must have turned it off the search". For anyone who suffer this: ask Claude to update memory with "a) 4.7 is buggy and tend to not use search, b) when unknown word/concept comes, always search"

@bcherny @GergelyOrosz This is really lame and must be painful for @bcherny - because this was easily reproducible, on any users. On me, Opus4.7 mapped "openclaw" to "opencode" without search. Not a big deal, you can add memory to "search each time when an unknown word comes in". However...

Claude just keeps regressing for me, day after day. I swear that until a few days ago, when Claude did not know something, it kicked off a web search, figured out, and answered. Now it just refuses to do the work that I pay for. It's like showing you the middle finger. Really?

The longest trip you will take is inside.

Basically

Pretty neat and probably right (?) approach. I mean, if I had to bet, I think the term "Mismanaged Geniuses Hypothesis" will stick to the community and will likely be shown (at least partially) to be true. Exciting times.

@morganlunt Nice feature, but two bugs stopped me from using it. - bedrock region: typing region "eu-central-1" not possible: char '-' stops the input. - once configured, then /login command disappear. I cannot come back to my Claude subscription unless edit the conf file...

Tired of manually writing config files and env vars to use Claude Code with Amazon Bedrock or Google Vertex? Me too. Just shipped a setup wizard that handles it for you. Bonus: Claude Code now notices if you have an older model pinned and suggests a newer one if you have access!

@vboykis Maybe we have hope. Claude Code/buddy pet gives me meaningless jokes every time it sees an interruption from the corner of his side of the terminal...

on programming joy and octocat vickiboykis.com/2026/04/06/on-…

@yoavgo Ah, good point. And I predict that just like the RTFM era, people won't read it and we will keep putting a lot of emphatic 'F' in RTF-anything.

RTFM era is over. we are now entering RTFC: Read The Fucken Code.

-93 bytes, one of the best Wikipedia line deletions I’ve ever seen.

I trained an LLM from scratch on pre-1900 text to see if it could come up with quantum mechanics and relativity. While the model is too small to do meaningful reasoning, it has glimpses of intuition. When given observations from past landmark experiments, the model can declare that “light is made up of definite quantities of energy” and even suggest that gravity and acceleration are locally equivalent. I’m releasing the dataset + models and leave this as an open problem to the research community. I also include what this project has taught me about intelligence in a mini essay linked below. 🧵(1/n)

The Wall fell before my time, but its strata turn up in the oddest places.

🇬🇧 From a Korean perspective, Germany is an enviably well-unified country. But 35 years on, the East-West line still snags on the smallest things. Not just election maps — blood glucose meters ship in mg/dL (West) vs. mmol/L (East), still a top return reason on Amazon.

한국에서 보면 독일은 부러울 만큼 잘 통일된 나라다. 그런데 35년이 지나도 동서 구분은 턱처럼 걸린다. 선거 지도만이 아니다. 혈당측정기 단위가 서쪽은 mg/dL, 동쪽은 mmol/L - 아마존 반품 사유 1위. 장벽은 옛저녁에 무너졌어도, 흔적은 쭉 이렇게 살아가는구나.

Everything is dead. I'm sick of it. Here's our answer: rip-grep.com

"AI can’t think, doesn’t have a mind and, in fact, is inherently untrustworthy." sigh... What was the term for such BS? 'not even wrong'?

@lossfunk It's a shame, because this was a good concept for a benchmark. At least, thanks to your benchmark, I've learned that Chollet will happily share anything that fits his narrative. That's a useful bit of information.