Speck: cognitive runtime lets a 4B model beat a naked 8B in speed and accuracy
Electronic-Space-736 · reddit · 2026-10-06
Speck is an open-source "cognitive runtime" that moves memory, attention, planning, evidence tracking, confidence and metacognition out of prompts into persistent, deterministic software — so the model is disposable but the cognitive state persists across restarts and model swaps. In its own benchmark (8 cases × 2 runs, identical sampling and 1,024-token cap), Speck on a shared qwen3:4b scored 15/16 at 7.9s/case and 5.7GB, beating a naked qwen3:8b (13/16, 13s, 6.7GB) with half the parameters. The author contrasts it with OpenClaw: OpenClaw is a strong agent runtime, Speck aims to be a cognitive runtime that makes the environment intelligent enough that the model doesn't have to be — e.g., constraining what semantic judgments truly need the model rather than upgrading to a bigger one. Self-assessed tables show Speck leading on architecture and observability, OpenClaw on tooling and production maturity.
More from coding & agent
- A Reddit agent postmortem: the ERP's note-wiping behavior dictated the guardrails — max_gladysh · 2026-10-06
- Replit CEO: someone left an AI agent running overnight and woke up to $10,000 of wasted tokens — amasad · 2026-10-06
- Replit's Shlomi Fruchter: MCP, Harness and Skills are absurd concepts doomed like prompt engineering — shlomifruchter · 2026-10-06
- One Prompt Dump, Full App: Developer Wowed by Spawn Agent's Single-Shot Build — TAbrodi · 2026-10-06
- After Five Years, MATHPETS Launches as a Language and IDE for Agent-Based Models — jessi_cata · 2026-10-06
- Reading the source of 7 LLM eval tools uncovered 13 scoring bugs, 6 fixes merged — maverick_man1111 · 2026-10-06