AI

Ep 97: Jev Creator on Decision Models, Benchmark Fatigue & Why the Labs Missed It

Read full article at source ↗

Based on the episode description, not a full transcript. Source evidence retrieved 05/11/2026, 15:04:25 UTC.

Briefings

Diogo Almeida, co-founder of TypeSafe AI and author of OpenAI's InstructGPT, joined Jacob Effron to discuss Jev, a 'system one' model launched in September that prioritizes structured decision-making over text. The conversation covers the technical challenges of building reliability, the 20-40% speedup coding agents achieve, and the distinction between LLMs and true AI agency. Key uncertainty remains regarding whether Jev's architecture could have been developed five years prior.

Key points

• Diogo Almeida coauthored OpenAI's InstructGPT paper and spent two years in stealth building Jev, a 'system one' model launched in September that returns structured decisions instead of text.
• Jev aims to solve the 'tyranny of the KV cache' in coding agents, potentially delivering a 20 to 40% speedup compared to standard LLM calls.

Context

Diogo Almeida, TypeSafe AI co-founder and author of OpenAI's InstructGPT, joined Jacob Effron to discuss Jev, a 'system one' model launched in September that prioritizes structured decision-making over text. The conversation covers the technical challenges of building reliability, the 20-40% speedup coding agents achieve, and the distinction between LLMs and true AI agency.

Limitations

Evidence is limited to a single podcast episode from November 2026. Specific launch dates for Jev are confirmed as September, but the exact timeline for its development and the precise reasons for the two-year development delay are not detailed. The episode does not provide data on model performance metrics or user adoption rates.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 05/11/2026, 16:00:28 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeigm32yw6xs4m4f2z6yeezryrp25nwab5e4jbftionb7pcldhzztw4