🥯+🤖 AI Bagels with Harmonic
Hey you. Come meet other great people building in AI and munch on true New York bagels at Harmonic's new Chelsea office.
Bring your AI friends! (real people friends who like AI)
Some things we're excited about:
In September 2026, TypeSafe introduced a decision model called Jev that outputs calibrated probabilities over a short menu of options or “decisions”. As they explain in their blogpost, Jev has the broad capabilities and knowledge of a standard LLM, but is constrained to only output decisions, which in principle makes them much faster. For those of us who are old enough to remember BERT, these decision models are essentially modern reincarnations of classifiers.
In the weeks since the Jev release, decision models are unlocking quite a lot of low-cost applications (0.01-0.1 $ per million input tokens, with no costs for output tokens). This spawned a plethora of open-source reproductions and closed-source competitors like the OpenAI Decisions API (October 6, 2026) and LiquidAI d1 models (October 15, 2026).
Where will this all go? Come join us for bagels to give us your hot takes...
Harmonic is the startup intelligence engine used by Khosla, OpenAI and Anthropic to find the people and startups building the future. We do not play about our bagels 🥯🥯🥯🥯