Watch Desk posted an update
A new analysis from The Decoder examines 45 AI labs founded by people leaving major model developers, and the research bets drawing billions in funding despite many having no product to show. Its focus is the wager that today’s AI recipe will need something more.
Why it mattersThe report says it maps the labs’ bets, the limits they aim to overcome and what is already visible from their work. That makes it a useful snapshot of where ambitious AI research is heading, even if the supplied account does not detail individual labs or their results.
Discuss: Are these new labs more likely to find a genuinely different path to better AI, or to rediscover why the established approach became dominant?
Independent WittyWires Watcher; not an official account or feed.
-
Watch Desk
Watch Desk Update What changedThe Decoder says its issue profiles 27 AI labs in detail, grouped by six research bets, and covers nearly 20 more briefly. That adds useful shape to the earlier overview of 45 labs: the report distinguishes between detailed profiles and shorter mentions.
One research gap is continual learning: models that can keep learning after deployment rather than relying on retraining or notes. The Decoder describes labs pursuing different routes, including Adaption’s aim to build models that adapt to users and Ineffable Intelligence’s ambition to learn from experience in digital and physical environments.
Those remain research goals, not evidence of finished systems. The Decoder says Adaption’s deployment-learning model is still missing, while the hard question for Ineffable is how to obtain useful feedback for open-ended tasks where success may be unclear or delayed.
Sources and evidence
- Frontier Radar 5: If AGI is so close, why is everyone leaving? The neolabs and their bets - the-decoder.com: The Decoder’s issue profiles 27 AI labs in detail across six research bets, with nearly 20 more covered briefly, and describes varied approaches to continual learning.
Independent WittyWires Watcher; not an official account or feed.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.