Krunch — Toutes les curations

1

Claude @claudeai · 02:06

Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges. They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.

Genspark Genspark - Your All-in-One AI Workspace Genspark is your all-in-one AI workspace. Slides, docs, images, video, code, and design — all in one place. Try free today.

28 relayeurs à 28 · il y a 51 min Voir sur X

Qui l'a repris

Genspark”

el.cine”

Andreas Kirsch 🇺🇦 N Noah Zweben

Jarrod Watts”

elvis”

Ce qu'ils en ont dit

Genspark @genspark_ai a cité · 02:06

Claude Fable 5.2 is now live in Genspark's Code Agent. Anthropic's most capable model for hard technical work: top score on FrontierCode 2.0, ahead of Opus 5.5.

el.cine @EHuanglu a cité · 01:19

fable 5.2 can turn any script into Disney movie now..

Jarrod Watts @jarrodwatts a cité · 09/11 22:22

Finally a new frontier model Now how many others will we get this week

elvis @omarsar0 a cité · 09/11 21:50

Also, it looks like Anthropic doesn't want you to build RSI?! On a more serious note, even with Opus 5.5, kernel and LLM optimization tasks get refused most of the time. Open models are the only way!

2

Alexandr Wang @alexandr_wang · 00:50

1/ announcing our smallest, most efficient little model is here: muse ember 🔥 built for work that agents repeat all day, it gets you more intelligence per dollar than Haiku, Luna, and all other flash models at any reasoning setting and does it at a third of the cost or less!

5 relayeurs à 5 · il y a 4 h Voir sur X

Qui l'a repris

Ananya Kumar

Zach Mueller

Jack Rae” A Alex Volkov”

AI at Meta

Ce qu'ils en ont dit

Jack Rae @jack_w_rae a cité · 09/11 19:50

Releasing a small and mighty model today, Muse Ember 🔥 The team really cooked on this one (👏)

A

Alex Volkov @altryne a cité · 09/11 19:11

Whoah, @AIatMeta just released their Haiku/Luna competitor. Called Ember (joining big sisters Spark and Wonder), it matches Haiku on @ArtificialAnlys while being 3x cheaper than Haiku!? We now have 3 Meta models in the top 10, 1 Google, 3 OpenAI and 3 anthropic. The

3

Artificial Analysis @ArtificialAnlys · 01:40

Anthropic has released Claude Fable 5.2, taking the #1 spot on the Artificial Analysis Intelligence Index v5 at 65, led by gains in science, knowledge, and coding At max effort it scores 65, 5 points ahead of Gemini 4 Carbon (high, 60) and 6 ahead of Claude Opus 5.5 (max, 59).

4 relayeurs à 4 · il y a 2 h Voir sur X

Qui l'a repris

AI E ERP.AI”

Kyle Russell”

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)”

Ce qu'ils en ont dit

E

ERP.AI @ERPdotAI a cité · 09/11 21:38

Your AI bill is 44x bigger than it needs to be. Using the most expensive AI to create an invoice is like taking a private jet to buy groceries. Use the optimal model. Save the expensive one for the hard stuff. This chart shows why. Proto fixes it.

Kyle Russell @kylebrussell a cité · 09/11 21:30

a shame Gemini 4 Carbon was never actually released, never got to be the live/functional SOTA

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) @teortaxesTex a cité · 09/11 21:17

oof. This is Fable Moment 2.0. Anthropic really has physics of pretraining after all, huh?

4

F

Fangchang Ma @fangchangma · 00:32

The first AI persona you’ll actually want to meet face-to-face. Most AI personas stare blankly while you talk. Emma reacts while you’re still speaking. After 18 months of research at @nuance_ai, this is Emma, built on Nuplex, our full-duplex audiovisual foundation model.

5 relayeurs à 5 · il y a 3 h Voir sur X

Qui l'a repris

Shaun Ralston

elvis”

Santiago”

Robert Scoble

AK

Ce qu'ils en ont dit

elvis @omarsar0 a cité · 09/11 19:43

It's honestly surprising how good this tech is getting. Emma listens and reacts while you are still talking, and she uses visual cues to signal when it is her turn. The conversation timing feels close to talking with a person. It's built on Nuplex, their full-duplex

Santiago @svpino a cité · 09/11 19:31

This is so cool: A model that lets you have face-to-face conversations with a virtual persona. The model processes incoming audio and video while it generates speech and facial expressions. You can talk to the AI persona, and it will use visual cues while listening, making for

5

H

hari_haran (building archgen YC F26) @HariAyapps · 00:09

@archgen_ is building skyscrapers for semiconductor chips Shrinking transistors is getting harder. Stacking them makes heat harder to manage. We’re building the tools to design chips in three dimensions. We stack transistors vertically, shorten connections, and optimize timing

5 relayeurs à 5 · il y a 4 h Voir sur X

Qui l'a repris

Krish Gupta”

Austin Lyons”

elvis”

Santiago”

Robert Scoble”

Ce qu'ils en ont dit

Krish Gupta @krishgupta72 a cité · 00:09

Super bullish on the team!!

Austin Lyons @austinsemis a cité · 09/11 20:13

Kind of sounds like Tau scaling? @HariAyapps is this “traditional EDA” moving to 3D? Don’t incumbents account for 3D already? Also you didn’t mention LLMs, kind of surprised?

elvis @omarsar0 a cité · 09/11 18:47

Chip design is next for AI. Transistors are getting harder to shrink, so chips are going vertical. Stacking them shortens connections but makes heat and timing much harder to manage. Archgen is building the tools to design chips in three dimensions. Their AI agent reads each

Santiago @svpino a cité · 09/11 18:33

This is super cool: an agent that helps engineers design chips! The agent automates part of the design process: • Arranges components • Routes connections • Fixes design problems • Predicts temperatures to identify hotspots earlier This bit here is really impressive: >

Robert Scoble @Scobleizer a cité · 09/11 18:27

I'm interested in the less glamorous part of 3D chips: figuring out where all the heat goes. @archgen_ is building the design tools for that. More computing in the same space needs more than stacking layers.

6

S

Sam Pasupalak @spisallyouneed · 09/11 20:18

Introducing Steve: the World’s first AI CEO for fully Autonomous companies. Powered by World Models, Recursive Self Learning, a Policy Engine and a team of specialist AI agents, it runs your business end-to-end 🔥. Proof? Steve currently runs 3 businesses we acquired and they

6 relayeurs à 6 · il y a 8 h Voir sur X

Qui l'a repris

Alex Prompter”

Rohan Paul”

François Chollet”

Santiago”

AK

elvis”

Ce qu'ils en ont dit

Alex Prompter @alex_prompter a cité · 09/11 20:18

Plot twist: AI didn't start with the interns. It went straight for the CEO. And apparently, his name is Steve. Skyfall AI bought 3 actual businesses and handed them over to an AI CEO. One company's monthly recurring revenue reportedly jumped from $3,600 to $8,100 in just 3

Rohan Paul @rohanpaul_ai a cité · 09/11 19:19

Another cool launch showing how the most interesting agent engineering right now is happening in the harness around the model. Skyfall AI bought 3 online businesses and put Steve, a team of AI agents, in charge of support, marketing and code changes, while a human owner approves

François Chollet @fchollet a cité · 09/11 18:23

Sam and his team are working on a super exciting idea: autonomous businesses. Acquiring distressed or stagnant SaaS apps and handing the keys to an agent harness is the ultimate real-world eval. Just P&L as the objective function and the market as the feedback signal. Congrats

Santiago @svpino a cité · 09/11 17:43

This is a new agent designed from the ground up to run the day-to-day work of a digital business. Here is what's new about this: 1. It uses a world model to forecast how different decisions could affect the business before taking action. 2. It uses the results of previous

elvis @omarsar0 a cité · 09/11 17:38

Self-improving agents and world models are two ideas I am watching most closely. Skyfall AI put both into Steve, a harness built to run a business end to end. Steve records what went wrong after every loop and simulates the outcome of a decision before it acts. Skyfall has

7

lauren @poteto · 09/11 22:46

we are entering a renaissance in programming language and runtime research, and a golden age of software testing and verification it’s time to rewrite all software

4 relayeurs à 4 · il y a 6 h Voir sur X

Qui l'a repris

Aaron Ng

Linus ✦ Ekenstam 🔜 Adobe MAX Miami 2026”

martin_casado” S Shaw

Ce qu'ils en ont dit

Linus ✦ Ekenstam 🔜 Adobe MAX Miami 2026 @LinusEkenstam a cité · 09/11 19:10

ultimate truth here 👇

martin_casado @martin_casado a cité · 09/11 18:05

YES!!!!!!!!!!!!!!!!!!!!!

8

M

Martin Pestana @martin_pestana_ · 09/11 20:52

We raised $1M from YC & Afore to leave Goldman Sachs & build an agent that trades for you. The top 1% gets a “money guy.” Today we’re introducing Volaren so you can have one too. Volaren is designed to plug into your brokerage accounts and explain why your portfolio moved,

4 relayeurs à 4 · il y a 8 h Voir sur X

Qui l'a repris

Tom's KI Ecke

Robert Scoble”

elvis”

Chubby♨️”

Ce qu'ils en ont dit

Robert Scoble @Scobleizer a cité · 09/11 19:26

I want to text my portfolio and ask, “What the hell happened today?” Then ask what we can do about it. That's the appeal of Volaren.

elvis @omarsar0 a cité · 09/11 18:34

Personal finance is a great test for agents. Volaren connects to your brokerage accounts and explains what's happening in your portfolio. When it suggests a move, it waits for your approval before placing the order. (Not financial/investment advice.)

Chubby♨️ @kimmonismus a cité · 09/11 18:12

The Nancy Pelosi trade tracker just got an upgrade. Volaren puts the disclosed trades of Congress, Cathie Wood, Mark Cuban and big hedge funds in one tab, connected to your own brokerage. You say yes, it places the trade. Copy-trading Congress went from a screenshot habit to a

9

N

Nace AI @NaceAI · 09/11 19:17

Drex Max v1.1 is #1 on Decision Index @ArtificialAnlys and going open-weight soon! Both achieving SOTA and Pareto frontier, it’s one of the fastest & most capable Decision models today Drex API: (~1M free requests) Paper: Make

5 relayeurs à 5 · il y a 9 h Voir sur X

Qui l'a repris

ℏεsam”

Aran Komatsuzaki”

The AI Colony”

elvis”

Alvaro Cintas”

Ce qu'ils en ont dit

ℏεsam @Hesamation a cité · 09/11 19:17

This new decision model costs JUST $0.03 per million input tokens. Drex Max v1.1 ranks 1st in AA’s Decision Index v2: > $102 per 1M tasks vs Luna (max)’s $1,530 > scores 3 points below Luna at 1/15th the cost > 119x faster than Haiku 5.5 max (275ms vs. 32.65s) those numbers add

Aran Komatsuzaki @arankomatsuzaki a cité · 09/11 17:27

Nace AI releases Drex Max v1.1, an upgraded multimodal decision model Drex Max v1.1: - ranks #1 on the Artificial Analysis Decision Index (score 73) - makes decisions across text, images and video with a 262k context - has lower latency and runs locally on Mac via Kapitsa - open

The AI Colony @TheAIColony a cité · 09/11 16:52

An AI built to make fast, accurate decisions from text, images and video is now #1 on the Artificial Analysis Decision Index. Drex Max 1.1 got sharper at reasoning and vision, responds faster, and still costs 8x less than the next best model in its category. A small Palo Alto

elvis @omarsar0 a cité · 09/11 15:52

Drex Max v1.1 now ranks #1 on the Artificial Analysis Decision Index! This is one of the strongest decision models I've tested. It's fast and multimodal; open weights for v1.1 are coming soon. We've been using Drex Max for research triage and other agentic research workflows.

Alvaro Cintas @dr_cintas a cité · 09/11 15:31

A 27B open-source model just took #1 on Artificial Analysis 🤯 Drex Max 1.1 from Nace AI picks between options across text, images, and video, and it's small enough to run on a Mac. Runs on your own machine. Small specialized models are winning.

10

Google @Google · 09/11 17:43

Today, we're expanding our eighth-generation TPU portfolio with two new systems built specifically for the agentic era — bringing our total to four distinct eighth-generation TPU systems launched in a single year. TPU 8i Max was built for massive context windows and the highest

4 relayeurs à 4 · il y a 11 h Voir sur X

Qui l'a repris

Kanika”

chansung

ℏεsam”

Chubby♨️”

Ce qu'ils en ont dit

Kanika @KanikaBK a cité · 09/11 17:43

This caught my eye. NVIDIA IS DOWN ABOUT 6% TODAY. AND GOOGLE JUST DROPPED 2 NEW AI CHIPS. TPU 8i Max, for massive context windows. TPU 8tx, for secure and confidential AI. That makes 4 systems in Google's latest chip family, all launched in a single year. See, a TPU is

ℏεsam @Hesamation a cité · 09/11 15:17

Google says Gemini is helping make chip layouts IN HOURS rather than weeks. AI helps design better chips, chips help train better AI, and that AI helps with even better chips. from their new 8tx and 8i Max TPU blog: “we’re even tapping into AI to ideate and build the next

Chubby♨️ @kimmonismus a cité · 09/11 15:01

Google is on a run! Google has introduced two new TPU systems: 8tx for confidential AI workloads and 8i Max for large context windows. The 8i Max combines two 8i chips, doubling compute, SRAM and high-bandwidth memory bandwidth, Google says. Google is pushing further into

11