{"slug":"jev-system","label":"Jev System","item_count":3,"day_count":3,"source_count":3,"first_seen":"2026-09-16T11:09:53+00:00","last_updated":"2026-09-21T23:09:20+00:00","generated_at":"2026-09-22T00:21:07.829675+00:00","sources":["langchain_blog","latent_space","simon_willison"],"days":[{"date":"2026-09-16","items":[{"title":"[AINews] Jev: a “System One Model” that only decides/classifies/routes/scores — >100x faster, >200x cheaper than small frontier LLMs","url":"https://www.latent.space/p/ainews-jev-a-system-one-model-that","source":"latent_space","type":"news","summary_1line":"congrats to TypeSafe!","sid":"487ee06b84c7fedd","published":"2026-09-16T11:09:53+00:00","editor_note":"TypeSafe AI's launch announcement: a decide/classify/route/score-only model, not a general chat model."}]},{"date":"2026-09-20","items":[{"title":"Jev: System One models for Prod, not God — with Diogo Almeida, CEO, TypeSafe AI","url":"https://www.latent.space/p/jev","source":"latent_space","type":"news","summary_1line":"The definitive Jev podcast with its lead creator.","sid":"9e45c053dcabfa34","published":"2026-09-20T18:51:38+00:00","sources":["langchain_blog","latent_space"],"editor_note":"CEO Diogo Almeida frames the design goal as production reliability over benchmark intelligence."}]},{"date":"2026-09-21","items":[{"title":"Jev introduces a new shape of LLM - System One, aka Decision Models","url":"https://simonwillison.net/2026/Sep/21/jev","source":"simon_willison","type":"news","summary_1line":"Last week TypeSafe AI unveiled Jev , their first example of a new category of model that they are calling \"System One models\" (I'm with Maggie Appleton, I think \"decision models\" is a better name for these). Jev is an...","why_it_matters":"Matches feed focus: eval.","sid":"012242336f7588da","published":"2026-09-21T23:09:20+00:00","editor_note":"Simon Willison's independent take, preferring the \"decision models\" name over TypeSafe's own \"System One\" branding."}]}],"editorial":{"tldr":"TypeSafe AI introduced Jev on Sep 16 as a new model category it calls \"System One\": small models built only to decide, classify, route, and score, claiming over 100x the speed and 200x the cost efficiency of small frontier LLMs for that narrow job. It is explicitly not a general chat or reasoning model.","stale":false,"whats_new":"The naming debate landed on \"decision models\" as the clearer term over TypeSafe's own \"System One\" branding, after CEO Diogo Almeida framed the Sep 20 launch podcast around production reliability rather than general intelligence.","why_it_matters":"A purpose-built model that's 100x faster and 200x cheaper than a small frontier LLM is a strong fit for the classify/route/score steps buried inside most agent pipelines — the exact spots teams currently overpay a general-purpose LLM to handle.","take_for_builders":"If a general-purpose LLM call in your agent pipeline is only doing classify, route, or score, benchmark Jev against it — the claimed cost and latency gap is large enough to justify a quick swap-in test.","beats":[{"kicker":"LAUNCH","tone":"launch","headline":"TypeSafe AI unveils Jev, a decide/classify/route/score-only model","summary":"Claimed to be over 100x faster and 200x cheaper than small frontier LLMs at these narrow tasks.","sids":["487ee06b84c7fedd"]},{"kicker":"RATIONALE","tone":"rising","headline":"CEO: built for production reliability, not benchmark intelligence","summary":"Diogo Almeida frames Jev's \"Prod, not God\" design goal in a launch-week podcast.","sids":["9e45c053dcabfa34"]},{"kicker":"COVERAGE","tone":"now","headline":"Independent take prefers \"decision models\" over TypeSafe's \"System One\" label","sids":["012242336f7588da"]}],"open_questions":["Do Jev's speed/cost claims hold up against a real agent pipeline's classify/route/score calls, not just TypeSafe's own benchmarks?","Will other labs ship a comparable decision-model category, or does this stay a single-vendor niche?"],"generated_at":"2026-09-22T00:30:00+00:00"}}