MACHINE LEARNING

→ View original post on X — @testingcatalog

MiniMax M3 live in Atomic Chat — HTML game

By

@testingcatalog

–

01 June 2026 17h30

MiniMax M3 is now live inside Atomic Chat 👀

Atomic tested M3 on a task to read a hand-drawn napkin sketch, write the game logic, build the UI, and ship a playable HTML platformer in one pass.

All this for $0.028 🤖 https://t.co/KfXqhytXNa pic.twitter.com/RyKjPioVH8
— 🚨 AI News | TestingCatalog (@testingcatalog) 1 juin 2026

MiniMax M3 is now live inside Atomic Chat. Atomic tested M3 on a task to read a hand-drawn napkin sketch, write the game logic, build the UI, and ship a playable HTML platformer in one pass. All this for $0.028.

→ View original post on X — @testingcatalog

M3: MiniMax’s latest model with 1M-token context and agentic reasoning

By

@testingcatalog

–

01 June 2026 17h30

M3 is MiniMax's latest model: 1M-token context, native multimodality, and agentic reasoning, built on their Sparse Attention architecture. Try it in Atomic Chat

→ View original post on X — @ylecun

Demucs produced at FAIR-Paris by @honualx

By

@ylecun

–

01 June 2026 17h30

Demucs was produced at FAIR-Paris by @honualx and collaborators.

→ View original post on X — @godofprompt

Agents need memory before tools, not expensive goldfish

By

@godofprompt

–

01 June 2026 17h09

We gave agents tools before we gave them memory.

That was backwards.

Now they can browse, code, email, book meetings, hit APIs…

…but still wake up tomorrow with no idea what they learned yesterday.

That is not an agent stack. That is a very expensive goldfish with… https://t.co/k1DqTtViOb pic.twitter.com/vH1kBgbqNt
— God of Prompt (@godofprompt) 1 juin 2026

We gave agents tools before we gave them memory. That was backwards. Now they can browse, code, email, book meetings, hit APIs… …but still wake up tomorrow with no idea what they learned yesterday. That is not an agent stack. That is a very expensive goldfish with

→ View original post on X — @runwayml

Runway announces London as European HQ and $100M AI investment

By

@runwayml

–

01 June 2026 16h09

Today we're announcing London as Runway's new European headquarters and our newest research hub focused on general world models. Over the next 18 months, we plan to invest $100M into the UK AI ecosystem, and that figure will more than double through 2028 as we scale our European

Paper addresses call for better benchmarks in private ML

By

–

01 June 2026 15h46

I'm excited since this paper addresses a call-to-arms in our #ICML2024 Best Paper with @florian_tramer and Nicholas Carlini, where we advocated for better benchmarks in private ML: https://
x.com/thegautamkamat
h/status/1603383883126669312
… 7/n

DP synthetic data benchmarks: hard even with large privacy budgets

By

–

01 June 2026 15h46

These tasks are hard/impossible to zero-shot, rather easy without privacy, but surprisingly hard even with large privacy budgets (ε = 100)! This room to grow means we can really measure progress made by new DP synthetic data benchmarks. 6/n

ContinuousBench prevents benchmark leakage and measures learning from data

By

–

01 June 2026 15h46

Benchmarks may leak. They might also measure something besides what we want to measure: how well does a model learn from data? ContinuousBench will be released periodically, preventing leakage, and tasks are chosen so that information is exclusively in the data. 4/n

New ContinuousBench tasks replace saturated DP-synth benchmarks

By

–

01 June 2026 15h46

3. Current DP-synth methods shouldn't perform too well: else, there's no room to distinguish new and better techniques. Classic benchmarks used for DP synth (e.g., IMDb, OpenReview) are effectively saturated. Our new ContinuousBench tasks (Geminon and News) satisfy 1-3. 3/n