Thor 雷神 ⚡️
@thorwebdev ·
Mexico City 🇲🇽💔 I really do like you, but the amount of time wasted in traffic here is insanity! 😵
Back
AI posts
AI posts
People and companies shaping AI. Filter your watched accounts, catch up on recent posts, or find the most liked and viewed.
Compiled by the Toolstash team · Updated
209 watched accounts · 2977 matching posts · Showing up to 3,000 recent stored posts. Counts are snapshots, not live X totals. This page refreshes every minute while visible. Last collected .
Marc Lou@marclou ·
They fixed it in 6 hours! 👏
Joshua Saxe@joshua_saxe ·
We need an equally memetic response video hyping nuanced thinking
kache@yacineMTB ·
Having kids was like a nuclear bomb going off for my career progression and I don't really care in the slightest
- kache
@yacineMTB ·
I don't think AI research or software engineering gets automated any time in a similar manner to mathematics soon. Like maybe more than a year out from that
There is just so much to it that.. depends. Like, they are tasks where understanding what you want is the bottleneck
Miles Brundage@Miles_Brundage ·
Still don’t think most people are aware of the fact that Hacktron could have taken images from *any* Slacks, not just OpenAI’s, a few months back
- kache
@yacineMTB ·
The best language models are complete dogshit. I can still push them to their limits. I'm so tired of how fucking bad they are
- kache
@yacineMTB ·
I don't understand how AI can be so bad at programming yet be so good at math
- kache
@yacineMTB ·
Seems extremely bad but also. If you hire people, human beings, you've been pwnd
Ethan Mollick@emollick ·
Artificial Analysis finally moves to closed benchmarks (goodbye GDPval-AA), and finds that open weights models are falling further behind than they had previously measured. Take with a grain of salt, but matches the general vibes that we haven't seen a Fable-class open model.
Theo - t3.gg@theo ·
🤔
Andrew Curran@AndrewCurran_ ·
I got this popup on my Pixel today while extracting text from an image.
- Theo - t3.gg
@theo ·
Slopalytics has been updated with the new data from Artificial Analysis. Changes the pareto line significantly
- kache
@yacineMTB ·
You can just miss flights. Like nothing bad will happen. It's really not a big deal. You can just do things
Mike Taylor@hammer_mt ·
They acquired the @levelsio of medicine
- Theo - t3.gg
@theo ·
It is easier to build features than it is to announce them
- Ethan Mollick
@emollick ·
Genie 4: "Scene: A Victorian high tea. Character: A mouse in appropriate garb." "A porcupine walks by" "The mouse grows wings" "The mouse flies to the teacup" This is realtime & I can walk around freely.
It is glitchy, but also impossibly advanced compared to a couple years ago.
Tibo@thsottiaux ·
Day 23/ Much requested iCloud Main & Calendar plugins. Just kidding, today is a reset. Still much requested though!
- kache
@yacineMTB ·
Just spoke to an insider
You got to go long middle aged dads in suburban Canada. Short American city dwellers
Greg Brockman@gdb ·
learn about php2shell and the general emerging security landscape:
Daniel San@dani_avila7 ·
OpenBMB just released MiniCPM-OCR, and the Jev model is really interesting!
A 1.3B model that extracts text, tables, and formulas from documents, paired with a dedicated System-1 Jev model built to reduce human review workload: it scores every result and flags only the parts that need a second look
It also achieves state-of-the-art results on document parsing benchmarks
Both models are open source, with support for: Ollama llama.cpp vLLM SGLang
The big win: no more re-checking every page by hand. Review only what Jev flags, and the human review workload drops significantly
Riley Brown@rileybrown ·
Can someone explain what AI plans will look like in 6 months? Who has a measured take on this? What will be the best plan for 200 or even 500$/ month in 6-12 months?
- Theo - t3.gg
@theo ·
Can someone make a bot that auto replies to every “why is this broken” t3 code tweet for me?
The answer is always the same: ‘npx t3 triage’
Artificial Analysis@ArtificialAnlys ·
Announcing Artificial Analysis Intelligence Index v5, featuring 2 new benchmarks and 3 upgrades. For the first time, the majority of the Index is built on private datasets
The Artificial Analysis Intelligence Index is our composite measure of model capabilities across tasks reflecting real-world usage today. v5 is our biggest upgrade since v4 launched in January. With v5, 60% of the Index is now based on evaluations with private test sets or held-out answers.
Alongside Intelligence Index v5, we’re launching the Artificial Analysis Progress Index v1, which tracks model capability over time on one continuous scale.
New to the Intelligence Index:
➤ FrontierCode 2.0: Cognition's newly-launched private benchmark of real-world coding tasks, graded on whether a repository maintainer would merge the agent's change
➤ Terminal-Bench-Science: 70 agentic scientific research tasks, run in a terminal environment
Upgraded:
➤ HLE-Diamond (replacing Humanity's Last Exam): a refined 1,000-question version of Humanity's Last Exam, split evenly between reasoning and knowledge questions
➤ AA-Omniscience v1.1: An updated dataset with saturated and public set questions removed, new questions added, and ambiguous or erroneous questions revised
➤ CritPt v2: a private, 40-question hard subset of CritPt v2, revised by the CritPt authors and run with a code-execution tool
Staying in the Index: AA-Briefcase, AutomationBench-AA, GDP.pdf and Terminal-Bench 4.0
Removed from the Index: GDPval-AA v2.1, SciCode, and AA-LCR v1.1
Key results:
➤ Claude Opus 5.5 (max) leads at 58, ahead of GPT-6 Astra (max, 57) and GPT-6.1 Sol (max, 55): Anthropic and OpenAI hold the top five models. Gemini 4 Argon (high) is next at 50
➤ The open weights gap widens from 3 points to 12: Kimi K3.1 Pro (weights not yet released) is the top open weights model at 46. Among released weights, DeepSeek V4.1 Flash 1017 (max) and GLM-5.3 (max) lead at 32 - 26 points behind the leader and below Claude Haiku 5.5 (max, 36)
➤ Cost per task varies by more than 13x among the leading models: of the 10 highest-scoring models shown, GPT-6.1 Terra (max) is the cheapest at $0.96 per Intelligence Index task and Claude Fable 5.1 (max) the most expensive at $13.04
- Mike Taylor
@hammer_mt ·
bring back golden gate claude
- Theo - t3.gg
@theo ·
If I have to go to a dashboard to use your product, I'm not going to use your product.
If anything, I'm more likely to tell Claude to rebuild your product.
roon@tszzl ·
seems reasonable to call the decision to not release GPT-2-XL safetyist slop separate from whether it made you an AGI-believer-elect or not in 2019. It was obviously true that the model wasn’t dangerous in any way; even the threats cited at the time were about spam
Alexandr Wang@alexandr_wang ·
this is about to be so many of your muses in singapore
ThePrimeagen@ThePrimeagen ·
1. should I try this 2. do you want a video on my experience
- ThePrimeagen
@ThePrimeagen ·
Ok, I think I have my first hill climb idea for an agent.
Kind of excited