Jef Newsom

@etechlead @unclebobmartin Meta-slogging!

@unclebobmartin But now it's a higher order slogging!

The Slog. We all know about the slog. We've been postponing a bit architectural refactoring because we know it's going to be a slog. But eventually the pressure builds and we heave a great sigh and begin the long arduous process of making a thousand dangerous changes and running the test suite as often as possible. Along comes the AI and suddenly the slog doesn't seem like such a big problem anymore. We just tell the AI to slog through, and twenty minutes later it's done; and it's right! And so off we go, confident that slogs are relegated to an ancient past. We'll never have to slog again! And then comes some deep systematic flaw that we must correct. And the AI simply cannot deal with it without hours of constant babysitting and monitoring. And there we are, slogging again.

@danpacary breath abated

New goal: 1T param inference MoE model on MacBook Pro Yes that’s 1 TRILLION here’s the deal. There are no rules.

@grok when are you going to get a proper CLI like all of the cool kids?

@elonmusk Can you say, “Oligarchy?”

We don’t live in a democracy if our elected leaders fail to implement the will of the people

@unclebobmartin blind allegiance, mostly.

Democrat politicians are now stuck defending two very unpopular issues. The defunding of DHS, and the opposition to voter id. I'm not sure how they get out of this hotbox unscathed.

@unclebobmartin Groundhog Day. That’s what happens in my experience..

These deep analytic dives into systematic failures burn a _LOT_ of tokens. It really has to think hard to work through the issues. It barely finishes before compaction. This implies something I think we've all known. There are problems that are too complex for the context window to hold. Once a problem exceeds the context window, I'm not sure what would happen. My approach would be to subdivide the problem into chunks that the AI could write a report about, so that it's conclusions would be available after the compression. This, however, simply postpones the issue. The final implication is that there is an upper limit of complexity beyond which the AIs cannot go. This must be true of humans as well, though we don't have context windows per se. Perhaps this explains why physicists have been stymied for over a century by the incompatibility of QM and GR.

It feels like some days Claude is a genius and other days he's mildly retarded. Still loveable, but frustrating.

@LeaderJohnThune And never underestimate your ability to pretend like you care

Starting today, we are going to have an important fight on the Senate floor. Polling shows broad support for all of the issues included in the SAVE America Act. But never underestimate Democrats’ ability to get on the wrong side of what the American people want.

@elonmusk @_kaitodev @garrytan @karpathy That assumes benevolent overlords, right? History might have different opinions about how power hoarding affects things. I'm actually very (VERY) optimistic about our AI future, but our current congress is case-in-point why we should be vigilant as all gitout.

@_kaitodev @garrytan @karpathy All jobs will be optional. There will be universal high income.

5 minutes ago, @karpathy just dropped karpathy/jobs! he scraped every job in the US economy (342 occupations from BLS), scored each one's AI exposure 0-10 using an LLM, and visualized it as a treemap. if your whole job happens on a screen you're cooked. average score across all jobs is 5.3/10. software devs: 8-9. roofers: 0-1. medical transcriptionists: 10/10 💀 karpathy.ai/jobs

@futureradar_FR Whoever thinks an AI doesn’t have an ego has clearly never coded with Claude or Codex.

🤖 L'IA VIENT DE METTRE LES POLITICIENS AU CHÔMAGE. Le Japon vient de basculer dans le futur. Le parti "Team Mirai" (composé d'ingénieurs et de codeurs) vient de faire un braquage aux dernières législatives avec 11 élus. Leur programme ? Confier les problèmes du pays à l'Intelligence Artificielle. Mieux encore : une autre formation politique a carrément nommé un chatbot IA (sous forme de pingouin virtuel) à la tête de son parti. Et honnêtement... est-ce une si mauvaise idée ? Une IA ne détourne pas d'argent public. Une IA lit réellement les dossiers avant de voter. Une IA n'a pas d'ego démesuré. Pendant qu'on débat sur l'utilisation de ChatGPT à l'école, le Japon l'envoie siéger à l'Assemblée. Prêts à voter pour un algorithme en 2027 ou dystopie totale ? 👇

@Will_Tanner_1 By that time, they had been fully employed for at least 5 years.

Ages of the Founding Fathers on July 4, 1776: -James Monroe, 18 -Aaron Burr, 20 -Alexander Hamilton, 21 -James Madison, 25 -Thomas Jefferson, 33 There is indeed a problem with people today being unserious, but to pretend that age itself is the problem is ridiculous

@elonmusk “With” is probably an important word here

Common sense. A simple test for immigrants is: would you enjoy working them. If so, they would be an asset to America.

@elonmusk This is a litmus test for the government. Is it by the people, for the people? Or is it something else entirely?

The Senate must enact the will of the people

@unclebobmartin Ideas are the gold rush. Might be fool’s gold, but it’s still a rush 😅

In the last three days I have: 1. Designed and implemented a complete JVM language that Codex believes (whatever that means) would be ideal for AIs to use regardless of what humans think about it. It compiles down to JVM bytecode. 2. Designed and build, from scratch, a wiki with it's own internal web server and fully described by Gherkin style acceptance tests. 3. Made significant updates to the computer strategy of the Empire game. 4. Produce the crap4java and mutate4java tools that I used to help build the wiki. 5. Conceived of and implemented the differential mutation strategy used in both my clojure and Java mutation tools. And for every one of those projects I implemented a strict TDD, ATDD, Crap, and Mutate workflow that forced coverage into the high 90s, kept Crap below 8, and split any files with more than 50 mutation sites. My poor laptop had all 16 (8 hyperthreaded) cores burning at 100%. The fan was raging the whole time. I was hopping from window to window overseeing the entire campaign. It was exhausting! Did that workflow slow the process down? Probably. Probably a lot. On the other hand all these projects maintained rigorous semantic stability, with all unit tests, and acceptance tests passing. I never ran the wiki until it was done. It worked first time. I never compiled a program with AIR-J until it was done. It worked first time. No bugs have been introduced into the Empire game (so far). And that, boys and girls, is a freaking miracle.

@MattWalshBlog If you view it as replacing, it does have the ramifications you suggest. If you view it as amplifying, then it is a remarkable tool. I’m able to use my creativity more productively than ever before. And it’s for things Claude or Gemini or Codex wouldn’t create on their own.

A weird thing is that whenever I post criticism of AI, the majority of the people agreeing with me seem to be leftists. It’s pretty much the only position I hold that seems to be more popular with the left than the right. It should be the opposite. My whole point about AI (especially AI in creative fields) is that it isn’t human, it doesn’t have a soul, and we cannot surrender our society to an unhuman soulless algorithm even if it makes our lives easier in some ways. It’s very strange that an argument predicated on the existence of the soul resonates with the left while conservatives tend to scoff at it.

@LeaderJohnThune Make it happen. No excuses. Be a leader.

Next week, I will bring the SAVE America Act to the floor. Senate Democrats will be forced to defend their outrageous positions on these issues and explain to the American people why common sense and the Democrat Party have parted ways.

@Austen :raises hand: on my near-term todo list. I need to get the low-level nailed first. Getting close. Then I'll be addressing higher-level semantics.

Who is building the better auth layer for agents? Because what we have currently is miserable

@elonmusk Bought and paid for.

Good question

@X the for you pop up, allowing me to filter by topic was brilliant when it was first released. The current version of it has completely lost the plot. Defaulting to everything selected is frustrating. I wouldn’t try to filter if I wanted everything. When I use it, I want to quickly filter for the thing I’m interested in in the moment, and then I’ll switch to something else later. That’s really useful. What it is right now is annoying.

@RaZ3us @karpathy Land, power, compute: buy now

"Forking an org" sounds like a cute developer feature until you realize it reduces the value of business execution to zero. If an entire company’s operational structure can be cloned with a single command, the corporate moat is dead. The only surviving monopolies won't be software startups; they will be the warlords who own the proprietary data, the energy grids, and the compute clusters. Everyone else is just building disposable templates.

Expectation: the age of the IDE is over Reality: we’re going to need a bigger IDE (imo). It just looks very different because humans now move upwards and program at a higher level - the basic unit of interest is not one file but one agent. It’s still programming.