Tessera
Folgen

Michiel Bontenbal

@mpbontenbal.eurosky.social

Complex beings as us humans can not be summarised in a few lines but I am here for #AI #climate #EUtech #EU_politics. And geopolitics, unfortunately. Lecturer in AI & IT. Some posts in Dutch. 🇪🇺

1836 Posts3056 Follower561 Folgt
mr. TIM@timkellogg.me · 17h ago

there’s a reason coding models have been a focus also, the biggest problem with RSI is power concentration. You can’t compete against Anthropic and OpenAI because they’re using tools you can’t even dream of now Astra is out and word is both have 2+ generations of models in the cooker

Tibo @thsottiaux X.com Astra was probably our biggest competitive advantage while it wasn't generally available. Since we've had it our productivity jumped so much that we shifted some of our plans 6 months ahead and will ship them at DevDay instead of mid next year.
Dare Obasanjo@carnage4life.bsky.social · 2d ago

Anthropic’s key innovation was the discovery that a harness was just as important as the model. This is why Claude Code essentially added a $1T to its valuation. The problem is a harness is not a moat. Everyone spent the first half of 2026 figuring this out. Now Codex, OpenCode, etc are good too.

Michiel Bontenbal@mpbontenbal.eurosky.social · 2d ago

3 days ago I explained to students that AI was very bad in solving maze games for ARC-AGI3, today Astra beats humans to it.

conputer dipshit@davidcrespo.bsky.social · 2d ago

most interesting gpt-6 post so far. it saturates ARC-AGI 3, which is something given that the previous best score was 30%. "we found it performing highly efficient, on-the-fly symbolic world modeling for each game and level." x.com/fchollet/sta... arcprize.org/blog/astra

François Chollet @fchollet ♥ 789 ⇄ 98 · 45m GPT-6 Astra represents a step-function change in model capability for interactive reasoning problems. It scores 66% on ARC-AGI-3 using our standard harness, and nearly 100% with a continuous conversation harness and custom compaction, at a cost of roughly $360 per game. In fact, the continuous harness version significantly outperforms our human baseline in action efficiency across almost all levels. When we examined the reasoning chains to understand how the model operates, we found it performing highly efficient, on-the-fly symbolic world modeling for each game and level. It goes as far as developing its own shorthand DSL to represent in-game situations -- essentially a game-specific algebraic notation. Overall, Astra exhibits symbolic modeling behaviors we had previously only seen with sophisticated harnesses -- so harness capabilities are increasingly shifting into the model itself. We see Astra as a major breakthrough in model intelligence. Read our post on Astra and what these results mean: arcprize.org/blog/astra
Andy Scollick@andyscollick.bsky.social · 7d ago

⚠️ On behalf of Hegseth and Trump, Elbridge Colby told NATO members that, essentially, in future Europe must fight its own wars (with Russia) without US involvement, but using expensive American-manufactured arms, unless it goes nuclear - in which case the US might get involved. That is #NATO 3.0 🧵

Bild
John Hanger@jrfhanger.bsky.social · last wk.

India is on track to reach its goal of 500 GW non-fossil fuel capacity by 2030. Crosses 300 GW level. Current capacities are: Solar: 165 GW Wind: 58 GW Hydro: 57 GW Biofuel: 12 GW Nuclear: 9 GW India will add more than 50 GW of solar just in 2026! oilprice.com/Latest-Energ... #energysky

India Surpasses 300 GW of Non-Fossil Power Capacity | OilPrice.com

India has surpassed 300 GW of non-fossil power capacity as rapid solar and wind growth pushes it toward its 500 GW target despite mounting grid constraints.

oilprice.com

Victoria Bosch@initself.bsky.social · last wk.

Are brains and artificial neural networks converging onto universal representations? There is a seductive idea making the rounds in NeuroAI / machine learning: train systems well enough, and they all converge on the same representation of reality (i.e. a unique world model). We have thoughts™ 1/n

The Umwelt Representation Hypothesis: rethinking Universality

Recent studies reveal striking representational alignment between artificial neural networks (ANNs) and biological brains, leading to proposals that all sufficiently capable systems converge on univer...

cell.com

Tom Aarsen@tomaarsen.com · 2w ago

📈 New blog post: Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers. As a practical example, I finetuned a ColBERT-style model for medical retrieval. 14.5 hours on one RTX 3090, and it beats every general-purpose retriever I could find. Thread 🧵

Bild