Top 10 AI posts for Nov 3, 2026
The ten AI stories on X that mattered most in the 48 hours to Nov 3, 8:04 AM ET, one post per story and one per person, ranked by how big the news is and how fast it was spreading. The list is re-ranked every two hours; this page keeps the latest one for the day.
Picked from the original posts of 30 accounts we follow for AI news, plus a scan of other recent AI posts. Each one was seen at least 5K times on X. Hot means more than 10K views.
-
Number 1. Rikiya Takehi @rikiyatakehi Hot: seen more than 10K times on X
MIT researchers had GPT-6 Astra directly edit 0.016% of a small Qwen model's weights, with no training, and its math improved up to 6x. Others say they reproduced it.
What it means for your workNothing to use yet, but it hints that models may soon tune or repair other models directly. Worth watching if you fine-tune your own models.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: https://t.co/bOWCcFhCVq
6.7M views 50K likes 6.5K reposts 2.1K replies Open post on X by @rikiyatakehi
-
Number 2. Praetor @FourVork Hot: seen more than 10K times on X
Anthropic is adding invisible watermarks to more Claude text from Nov 3. It works by nudging word choice, so retyping the text by hand does not remove it.
What it means for your workAssume text drafted with Claude can be flagged as AI by authorized checkers. Edit important drafts in your own words, and check client or school rules.
That's not how AI watermarking works. Claude doesn't insert invisible characters, metadata, or extra tokens into your text. it uses a version of Google's SynthID, which subtly changes how the next token/word is selected, leaving a detectable statistical pattern in the wording. So you'll need to rewrite the text to get rid of it.
262K views 1.8K likes 50 reposts 83 replies Open post on X by @FourVork
-
Number 3. DeepSeek Harness @DeepSeekHarness Hot: seen more than 10K times on X
DeepSeek Harness added Creator Mode, which turns a plain description into plugins you can use inside the app, from a design workspace to a sandbox game.
What it means for your workYou can describe a small internal tool and have it built inside the app. Try it on a simple helper before trusting it with real work.
Your Harness. You Define It. Meet Creator Mode in DeepSeek Harness. Turn ideas into plugins that you can use, from a design workspace to a sandbox game or a 3D LLM visualizer. What will you create? https://t.co/MlXHIyyA1o
2M views 23K likes 1.9K reposts 636 replies Open post on X by @DeepSeekHarness
-
Number 4. unusual_whales @unusual_whales Hot: seen more than 10K times on X
Per Bloomberg, Goldman Sachs restricted most staff from AI coding tools like Claude Code over productivity and risk concerns, keeping them mainly for engineers.
What it means for your workIf staff outside engineering run coding agents, set clear rules on who can use them and on what data, before a big client or regulator asks.
BREAKING: Goldman Sachs has restricted most staff from using AI tools designed for coding due to productivity and risk concerns, per Bloomberg.
640K views 7.6K likes 498 reposts 217 replies Open post on X by @unusual_whales
-
Number 5. Chenru Duan @chenru_duan Hot: seen more than 10K times on X
Deep Principle released Hydrogen-1.0, an open-weight science model built on GLM-5.3. It scores 38.3% on FrontierScience-Research vs GPT-6 Astra's 33.3%.
What it means for your workMostly for research labs. If you do science or R&D work, it is a free model you can adapt to your own data.
We’re introducing Hydrogen-1.0, our first open-weight model for scientific research. We post-trained GLM-5.3 (@Zai_org) on our internal scientific data, lab records, and discovery trajectories. It scores 38.3% on FrontierScience-Research vs GPT-6-Astra’s 33.3% and scores 93.0% on GPQA-Diamonds. We believe science should be open, collaborative, and democratized, rather than proprietary. Only open weights let researchers inspect, adapt and build on the unprecedented speed of AI growth. There are many recent models named by chemical elements, but we believe Hydrogen-1.0 is the first aiming…
1.3M views 5.4K likes 434 reposts 139 replies Open post on X by @chenru_duan
-
Number 6. Guillermo Flor @guilleflorvs Hot: seen more than 10K times on X
Revolut moved about 2,000 engineers to a self-hosted GLM model on 56 Nvidia B200s, at a blended 30 cents per million tokens, roughly a tenth of its old frontier lab bill.
What it means for your workHeavy AI coding users can compare open models against their current spend. For most small teams, a hosted open-weight API is the easier first test.
BREAKING: Revolut moved ~2,000 engineers onto self-hosted GLM for coding and pushed Claude Code and Codex to the exception list. Per David Abitbol, who leads AI ops there: 56 Nvidia B200s, a blended 30 cents per million tokens, roughly a tenth of what they paid Anthropic and OpenAI. Engineering lead time improved 70%. Here's the part that matters: the gains showed up when tokens stopped being rationed. Engineers quit saving AI for the hard tickets and pointed it at everything. Frontier models become the exception, open weights become the default. That's a pricing problem for every lab…
275K views 3.7K likes 317 reposts 139 replies Open post on X by @guilleflorvs
-
Number 7. Tibo @thsottiaux Hot: seen more than 10K times on X
OpenAI is expanding the frontier models that power dots, including GPT-6 Astra and GPT-6.1 Terra, with more coming. Terra brings speed.
What it means for your workIf you use dots agents, expect faster and cheaper runs on simpler tasks as Terra joins. Nothing to change today.
Fellow resetters, heads up that we're working on expanding the family of frontier models that power dots, including GPT-6 Astra and GPT-6.1 Terra and some others in the future. Terra brings speed and a nice personality upgrade, and will help us bring dots to more people.
1.7M views 11K likes 309 reposts 1.8K replies Open post on X by @thsottiaux
-
Number 8. SemiAnalysis @SemiAnalysis_ Hot: seen more than 10K times on X
SemiAnalysis released GDPx, an agent benchmark of 206 expert tasks across 11 valuable fields. Even the best model scores just 12%.
What it means for your workAgents still miss most expert-level tasks end to end. Keep a person reviewing high-stakes output, and note cheaper models score close to the leader.
Today we’re releasing GDPx, a benchmark for measuring agentic capabilities across the most valuable domains. GDPx is a true frontier-level benchmark: even the best model scores just 12% (1/7)🧵 https://t.co/RNnkcuUkih
1.7M views 3.3K likes 310 reposts 355 replies Open post on X by @SemiAnalysis_
-
Number 9. Steve Weis @sweis Hot: seen more than 10K times on X
Anthropic released FlashProve, an open source library for provable inference, tested on Llama 4 Maverick, a 400B-parameter model. Code and a blog post are public.
What it means for your workDown the road, you may be able to verify which model actually answered you when you pay for a premium one. Nothing to act on yet.
Anthropic is releasing an open source library, FlashProve, for provable inference. It was measured on Llama 4 Maverick, a 400B-parameter mixture of experts model. Code: https://t.co/1ILw5Zf8Lh Blog post: https://t.co/YROBN2PThF
52K views 482 likes 62 reposts 34 replies Open post on X by @sweis
-
Number 10. Pierre Stock @PierreStock Hot: seen more than 10K times on X
Mistral released ML4 as open weights, with frontier skills on code agents and cyber, smaller and more efficient than its preview three weeks ago, plus a technical report.
What it means for your workAnother strong open model to self-host or run through a provider, useful if you need control over data or costs for coding agents.
ML4 is now open-weight! Frontier capabilities on code agents & cyber, way more efficient and compact than its Preview version three weeks ago (!) and cherry on the cake: technical report telling its epic backstory. Check it out: https://t.co/TOOfN9Pob2
74K views 510 likes 89 reposts 20 replies Open post on X by @PierreStock
Need help putting this to work? FriendlyExperts fixes apps built with AI and sets up AI for small businesses, at a fixed price.
The Top 10 in your inbox every morning
The ten most talked-about AI posts in plain English, with what each one means for your work and a link to the post. No scrolling through social feeds.
You're on the list.
You're signed up for the daily Top 10 email.