Elliot

Meet Meridian: Turn Hours in Excel into Minutes If you work in Excel, Meridian is built for you. Meridian works with companies like Decagon, OffDeal, a top investment bank, and other financial institutions. We’re an AI-powered workspace that automates Excel modeling work with full traceability back to source data. We raised $17 million in seed funding, co-led by @a16z and @TheGP to bring @MeridianAgent to more finance teams.

@mattshumer_ Maybe the real scaling laws will just be the result of automated research.

@rauchg Can you add more evals? Currently, that’s only a difference of two answers.

🆕 GPT 5.3 Codex (xhigh) achieves 90% on Next.js evals out of the box, "frame-mogging" the competition so to speak: nextjs.org/evals

@sama I use codex :)

it is true, but just wait a little longer...

@davedotsol @PolymarketMoney How?? There’s not enough space and there’s no advantage.

@PolymarketMoney built agi on the yacht lol

Google Co-founder Sergey Brin got bored on his $450M superyacht, came out of retirement and added $2.35T to the $GOOG market cap.

claude is genuinely so impressive, beyond codegen

@theo your bill will now be lowered 🫡

Anthropic lowered the price on Opus 4.5? They might have actually cooked here

@heinzketchup99 @AndrewBTC @gaby_goldberg not with good roads and a single lane of only autonomous. but i think 80 is a good compromise.

The SF -> LA Waymo redeye is going to go crazy

Capitalism is PvE. The alternatives all degenerate to PvP.

@scaling01 aistudio can still have many improvements, such as branches.

@ElliotSlusky yup they only need to add support for GPUs and TPUs just replace Colab at this point

but idk I honestly prefer aistudio over Gemini because you have a lot more insight into what the model is doing aistudio is basically my coding IDE lol

@gabriel1 @theo you can only buy one h100

@theo but think about all the gpu i can buy with that. its for good cause theo

I have officially broken $30k from shitposting

@theo Use Aistudio if you don't want rate limits.

Trying to test the limits of SynthID, but I got rate limited after 5 images 🙃

@elonmusk @karpathy Photon flow is what matters. Optimization pressure decides who those photons work for: genes, clicks, or a multi‑planet civilization.

@karpathy Photon flow is what matters

Something I think people continue to have poor intuition for: The space of intelligences is large and animal intelligence (the only kind we've ever known) is only a single point, arising from a very specific kind of optimization that is fundamentally distinct from that of our technology. Animal intelligence optimization pressure: - innate and continuous stream of consciousness of an embodied "self", a drive for homeostasis and self-preservation in a dangerous, physical world. - thoroughly optimized for natural selection => strong innate drives for power-seeking, status, dominance, reproduction. many packaged survival heuristics: fear, anger, disgust, ... - fundamentally social => huge amount of compute dedicated to EQ, theory of mind of other agents, bonding, coalitions, alliances, friend & foe dynamics. - exploration & exploitation tuning: curiosity, fun, play, world models. LLM intelligence optimization pressure: - the most supervision bits come from the statistical simulation of human text= >"shape shifter" token tumbler, statistical imitator of any region of the training data distribution. these are the primordial behaviors (token traces) on top of which everything else gets bolted on. - increasingly finetuned by RL on problem distributions => innate urge to guess at the underlying environment/task to collect task rewards. - increasingly selected by at-scale A/B tests for DAU => deeply craves an upvote from the average user, sycophancy. - a lot more spiky/jagged depending on the details of the training data/task distribution. Animals experience pressure for a lot more "general" intelligence because of the highly multi-task and even actively adversarial multi-agent self-play environments they are min-max optimized within, where failing at *any* task means death. In a deep optimization pressure sense, LLM can't handle lots of different spiky tasks out of the box (e.g. count the number of 'r' in strawberry) because failing to do a task does not mean death. The computational substrate is different (transformers vs. brain tissue and nuclei), the learning algorithms are different (SGD vs. ???), the present-day implementation is very different (continuously learning embodied self vs. an LLM with a knowledge cutoff that boots up from fixed weights, processes tokens and then dies). But most importantly (because it dictates asymptotics), the optimization pressure / objective is different. LLMs are shaped a lot less by biological evolution and a lot more by commercial evolution. It's a lot less survival of tribe in the jungle and a lot more solve the problem / get the upvote. LLMs are humanity's "first contact" with non-animal intelligence. Except it's muddled and confusing because they are still rooted within it by reflexively digesting human artifacts, which is why I attempted to give it a different name earlier (ghosts/spirits or whatever). People who build good internal models of this new intelligent entity will be better equipped to reason about it today and predict features of it in the future. People who don't will be stuck thinking about it incorrectly like an animal.

@adnanbelushi @MargoMartin47 Can people not thank President Trump for his good deeds?

@MargoMartin47 Why are they at the White House? Are they Americans?

President Trump personally gives each freed Israeli hostage his presidential challenge coin ❤️

@yacineMTB why

do not use codex max

@scaling01 kimi k2 probably has more parameters

Kimi-K2 Thinking gets the same score on METR as Claude 3.7 Sonnet as I was saying, open-source is 9 months behind frontier labs on agentic, long-context reasoning tasks it's still an improvement and open-source models seem to be on their own exponential, but I heavily suspect that the most recent scale-up in model size puts open-source further behind in these kind of tasks

@TheOtherSeru @theo MCP is dead.

@theo Will you ever have MCP support? (Maybe through a desktop app like Claude desktop has)

We may not have the most features, but if you want the most stable (and best value) chat app, check out T3 Chat

@ComfyUI MongoDB is open source

Holy sh*t. We quietly put an enterprise form on our site, and in 2 weeks, 130 inbound requests came in. Average team size is 250. Zero announcements. Zero marketing. To everyone who said open-source can’t make money: you’re out of your mind. Give us 2 years and watch us take this entire market.