Non-neural assistant CAEL 0.3.6 ties Qwen3-0.6B on 400-question conversation test
Silver_Employ2617 · reddit · 2026-10-28
A developer released CAEL 0.3.6, an experimental assistant running with zero neural networks, transformers, or external LLMs — combining symbolic solvers, counted language statistics, retrieval, and automaton components. The update focuses on conversation reasoning: applying later corrections to stored facts, flagging conflicting reports, combining values across messages (e.g., "bought 12 eggs, used 7" → 5 left), and refusing to state unstated prices.
On the author's local 400-question phrase-graded conversation test:
- CAEL 0.3.5: 53/400
- CAEL 0.3.6: 122/400
- Qwen2.5-0.5B-Instruct (fp32): 102/400
- Qwen3-0.6B (fp32): 121/400 — effectively a tie with CAEL 0.3.6
The author stresses this is a narrow local evaluation, not an independent leaderboard, and CAEL still struggles with references, planning, and unsupported questions. The 563 MB release passed 172 focused tests, 129 subtests, and 12 clean-install checks; source-available under a non-commercial license, with commercial use requiring a separate license. AI coding assistants (Codex) contributed to development.
More from coding & agent
- Snowflake Doubles Down on Agentic RL: Overfit First, Then Scale to 65.7 on SWE-bench — StasBekman · 2026-10-28
- Hook-level agent security wins: Microsoft, Anthropic and OpenAI all shifted from gateways to runtime hooks — jacob_posel · 2026-10-28
- Parsing Claude's Raw SSE Stream into a Forensic Timeline of Agent Tool Calls and Diffs — Additional-Ladder-24 · 2026-10-28
- Supabase launches Compute in private alpha: run AI agents and long-running jobs next to your Postgres data — dshukertjr · 2026-10-28
- Agency Owner Predicts Agents Go Mainstream in 2027 — Built Into Products, Not Self-Set Up — tjrobertson-seo · 2026-10-28
- Zero-AI-at-runtime design generator hits run 8, new aesthetics expected by run 10 — Afinetheorem · 2026-10-28