Source
· Daily Brief
AI Daily Brief — 8 May 2026
Sequoia Capital published AI Ascent 2026 as its annual state-of-the-field essay. Sakana AI released Twell — sparser, faster, lighter transformer language models. BAIR proposed adaptive parallel reasoning as the next paradigm in efficient inference scaling, and METR reviewed the Anthropic Risk Report.
Top stories
- Sequoia Capital: AI Ascent 2026. Annual state-of-the-field essay from a top investor. via Sequoia
- Sakana AI: Twell — sparser, faster, lighter transformer language models. New architecture release from Sakana. via Sakana AI
- BAIR Berkeley: Adaptive Parallel Reasoning — the next paradigm in efficient inference scaling. Academic paper proposing a new test-time-scaling formulation. via BAIR Berkeley
- METR: Review of the “Risks from automated R&D” section in the Anthropic Risk Report. Third-party analysis of Anthropic”s safety reporting. via METR
- Dwarkesh Patel: David Reich on why the Bronze Age was an inflection point in human evolution. Long-form podcast. via Dwarkesh Patel
Quiet corners
OpenAI, Anthropic, Google DeepMind, Mistral, Meta (Llama), DeepSeek were absent.
By the numbers
- 9 stories tracked across 7+ sources
- Most-mentioned model: Twell
- Most-mentioned lab: Sakana AI
- Notable absences: OpenAI, Anthropic, Google DeepMind, Mistral, Meta, DeepSeek
Compiled by AI Feed”s editor from all 9 headlines published on 8 May 2026.