James Kelly

guitar, lit, cs, healthcare

[James Kelly @heartsalve]

Helen Frankenthaler, Ramblas (1987-1988)

[James Kelly @heartsalve]

@fchollet is it just an optimization, or is it provably not learnable from a continuous distribution across the token space w sgd? are tokens fundamentally asymbolic in some way?

[James Kelly @heartsalve]

Shameful. I have always believed that good science was predicated on strict ethics.

[Gautam Kamath @thegautamkamath]

How do you tell if authors fabricated their research results? Sometimes: they just tell you. Read, e.g., this caption carefully.

Another banger of a post by my friend Nihar B. Shah on his blog, where he found at least 20+ such #ICLR2027 submissions:
https://t.co/TBKmnMjTiy

[James Kelly @heartsalve]

@unconed next console gen getting an 8b model, $1500

[James Kelly @heartsalve]

boo llm syntax: if (found++ == 0)

[James Kelly @heartsalve]

@DKThomp But they’re not lonely though. We all have jobs and talk to people all day man.

[James Kelly @heartsalve]

@ThosVarley low entropy

[James Kelly @heartsalve]

@jshguo beautiful work!

[James Kelly @heartsalve]

@francoisfleuret By your definition, mycelium are more conscious than current AI models.

[James Kelly @heartsalve]

@francoisfleuret The results are in, and it's even better than we thought.

[James Kelly @heartsalve]

@VictorTaelin I'll take that bet!

[James Kelly @heartsalve]

@cloneofsimo doctors, nurses, phramacists, car dealerships, traders, teachers at all levels, students who abstain from using it and are effectively penalized for it, human resources people, psychologists, botanists, biologists....

[James Kelly @heartsalve]

@willdepue No robot will ever function as well and completely as a single living organism.

[James Kelly @heartsalve]

@trq212 I agree with this. I do think it’s important to keep an ongoing document of what everything does, why, what’s tried, what’s on the table, what’s tested, what’s untested, etc. Claude.md and memory aren’t enough for this/bring too much noise to signal.

[James Kelly @heartsalve]

All engineering jobs this way now?

[voxium @v0xium]

I am done with this shit. It is over. The state of engineering right now is horrible. It has been half a month since I started a new role at a big company. Nobody knows anything here. The specs, code, tests, PRDs, tickets, resolution of those tickets, reports, etc., everything is made by Claude Code. Nobody on my team likes this. They are being forced to ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck, so why are we slow? People are working 12 to 13 hours a day just to press enter. Nobody is reading anything. Humans in corporate are doing nothing on their own. Everyone, literally everyone, from an L1 to an L7 engineer here is doing the same thing. Talk to Claude. There is no sense of victory. Nobody is resolving bugs. In reality, nobody is thinking anymore. Everything is done by LLMs. It is so soul-sucking. I would not mind it, to be honest, if we were at least given the time to check out the code and see what is going where. But no, the goal is to just ship. No matter what happens.

[James Kelly @heartsalve]

@VictorTaelin Show me a snippet. Is it beautiful?

[James Kelly @heartsalve]

@RosuGrigore I would prefer a language that does not allow nested assignment.

[James Kelly @heartsalve]

@poetengineer__ that's a really interesting point about universailty and how to represent every face you represent noface.

[James Kelly @heartsalve]

@_friendsthat @Chase This happens in other countries too. In Spain we were asked to provide tax documents for any deposit over 10k.

[James Kelly @heartsalve]

The hardest problems in the universe are not math problems.

[James Kelly @heartsalve]

vibecoding is terrifying:
"Worth noting how close that came to shipping: it produced a plausible table, and only looked wrong because two different architectures agreeing to four decimals is impossible."