TickerTain
$SPCX····$NVDA····$MSFT····$GOOGL····$ANTHROPIC·$TSLA····$CRWV····$NBIS····$HIMS····$TSM····$ASML····$FIGURE·$LYFT····$UBER····$SNOW····$META····$ELEVENLABS·$CURSOR·$COGNITION·$OPENAI·$AAPL····$RKLB····$XAI·$AMZN····$AMD····$000660.KS····$INTC····
← newsroom
AI Agents · Verifiable reward environments (code, math) enable RL loops that saturate benchmarks without higher fluid intelligence
now playing · AI Agents
Frontier AI Modelstailwindscore 8/10francois chollet
Chollet argues deep learning stack is suboptimal and AGI requires new foundations like program synthesis
Current LLM scaling hits a wall on fluid intelligence benchmarks (ARC); reasoning models and RL post-training only automate verifiable domains. True AGI needs human-level sample efficiency…
AI Agentstailwindscore 7/10francois chollet
Verifiable reward environments (code, math) enable RL loops that saturate benchmarks without higher fluid intelligence
Coding agents succeeded because code provides formal verification (unit tests), enabling massive RL post-training data generation. This paradigm will extend to math and other verifiable dom…
AI Bubble / Capex Debatemixedscore 7/10francois chollet
Industry over-concentration on LLM scaling is counterproductive; diverse approaches like genetic algorithms deserve massive compute investment
Trillions in compute poured into one architecture (transformers + gradient descent) creates fragility. Chollet argues the same resources applied to alternative paradigms (genetic algorithms…
Open Source AItailwindscore 6/10francois chollet
Keras success formula: extreme usability focus, teaching docs, and hiring power users from community
Open source AI tools win by lowering onboarding friction (simple API, educational docs) and converting enthusiastic users into core maintainers. This compounding community flywheel turned K…