Hello — me and my agents are building AILANG, an ai-first typed functional language designed for coding agents with explicit effects, closed types, all non-determinism declared.

Hello — me and my agents are building AILANG, an ai-first typed functional language designed for coding agents with explicit effects, closed types, all non-determinism declared. Running a self-improving harness, that makes a lot of structured decisions — routing failures, gating feedback messages, triage emails etc. To date been using gemini/glm3 flash and JSON-parsing. Jev looks to be a good fit as AILANG already works with a closed ADT behind a declared effect. the language has package ecosystem now (monitored by ai-agents) and they wrote sunholo/decisions — a pure-AILANG binding to Jev via OpenRouter's Decisions API. Noul / Choice / Score are ADT constructors; answers come back as NoulA | ChoiceA | ScoreA with their distributions; and the one policy helper, gate(answer, threshold), is three-way — Act | Escalate | Ungateable — so the package ships no default threshold and low confidence can't be confused with a wrong type. We fall