Files
hermes-brain/ideas/passepartout-economics/time-estimates.org
Hermes 303e8c6306 Convert cross-references from [[id:uuid]] to [[file:name.org]]
All 117 inter-node links now use [[file:node-name.org][title]] format
which renders as clickable hyperlinks in both Emacs (C-c C-o) and
web-based org renderers (Gitea, GitHub). Each node retains its :ID:
UUID property for Emacs org-roam database features (backlinks,
capturing, node-find).

Prev format: [[id:uuid][title]] — Emacs only, dead text on web
New format:  [[file:name.org][title]] — works everywhere
2026-05-21 19:40:54 +00:00

1.3 KiB

Development Velocity and Timeline Estimates

At the observed velocity (v0.4.0 to v0.7.2 in a single session), the agent writes code and the symbolic engine verifies it at a cycle measured in minutes. The bottleneck is not coding speed — it is LLM API latency, ACL2 verification time, and human review of the 5% of edge cases Screamer flags.

To v1.0.0 (neurosymbolic maturity, ~4,500 lines): ~80 cycles, 3-5 weeks, ~2-3 hours of human review.

To self-driving Lisp Machine (Logos + Stoa hardware, +~6,000 lines): ~60 cycles, 2-4 weeks. The microcode must be loaded onto physical hardware and benchmarked, adding seconds per cycle.

Full Stoa (editor, browser, shell, Qt integration, ~3,500 lines): ~30 cycles, 2-3 weeks.

Total from today to full Logos + Stoa + Agora triad: 3-6 months. Most of that time is spent on design decisions and protocol specification, not on code.

The system writes the code. The human makes architectural decisions and reviews the 5% ambiguous rules.

See also: Self-driving Lisp Machine, Sufficiency flip, Investment thesis