Do anybody still download @WordPress themes on @ThemeForest and host it? Those were the days

https://t.co/vnq4Zhvtf1 This is pure gold for understanding GPUs

Here’s also the numbers from Jeff on “Numbers Every Engineer Should Know” in case you need it

I stumbled up on a GitHub repo while reading some docs on optimizing LLM inference which talks about “Numbers Every LLM Developer” should know inspired by @JeffDean — here’s the repo should be helpful if you’re trying to serve your own model https://t.co/gWIL0ZlRc6 It’s a bit

𝕏 algo seems really on point. Gives me exactly what I need when I need. Almost like some kinda telepathy

Yesterday we shared our @Alibaba_Qwen's 3Guard benchmark results. Today, here’s the NotebookLM video that walks through what those numbers actually mean how context length impacts LLM safety, and why 7k-token chunks hit the sweet spot.

I like the @LangChainAI docs so much better now. @hwchase17 is there a way to remove inbuilt tools without getting too internal in the deep research agent like read and write etc. I’m using that for a use case where that’s not needed. It sometimes makes those tool calls when not

Just in case you’re wondering us-east-1 was a DNS issue

If you want to give it a read here’s the paper:

we’re not “halfway to AGI” - we’ve built something completely alien. superhuman at some tasks, worse than a kid at others. the shape is totally jagged. Also they’ve defined all the terms everyone keeps mixing up - “recursive AI” (AI building better AI) vs “superintelligence” vs

GPT-5 scores 58% overall - sounds decent but look at that chart - it’s absolutely cracked at math and knowledge but literally can’t do spatial reasoning or basic planning. GPT-4 scored 0% on reasoning.

they broke intelligence into 10 pieces based on real psychometric science (the same stuff psychologists use to test humans). Math, reasoning, memory, visual processing, all of it.

okay so i’ve been going through the paper “A Definition of AGI” by Hendrycks et al. and it’s genuinely the first paper that actually DEFINES what AGI means. like we’ve all been arguing “when AGI” for years but nobody agrees on what it even IS. This paper defines: AGI =

It’s surprising to consider that in the future, many autonomous agents might rely on the cloud, and if AWS experiences downtime, they would also cease to function, resembling a dystopian technology glitch. #AGI is just AWS backend.

Indian devs out on diwali and @awscloud is down. Surely one has nothing to do with the other. Right? Right?

Take a loot at this thread! If you’re building on LLMs (most of us are 😌) get your LLMs protected from various threats.

I think we’re all blessed to have @AnthropicAI @claudeai 🥹

@AnthropicAI

Only me or @AnthropicAI ERROR agent - Anthropic API call failed: Error code: 529 - {'type': 'error', 'error': {'type': 'overloaded_error', 'message': 'Overloaded'}}

A Friend: What do you do as an AI Engineer? Me: I spend most of my time reading @LangChainAI docs and slide into @AnthropicAI Claude's DM. 😌