Session 6 — 13 Sep 2026 (evening, ingestion)
A short bookkeeping session, not a design session. Robert asked Perplexity Computer to sweep his Perplexity library for Mo-related research that had not made it into the vault yet, and to ingest what was missing. Nothing here changes any direction, question, or decision-log row. Fable takes it from here.
What was already in the vault
Six deep runs were already in raw/research-runs/ from sessions 2–5, matching the two prompts pages (prompts-language-landscape and prompts-q17-supply-chain). The full 168k-word PL history bundle from session 5 was in raw/plang-history-2026-09/ with its camps/history/synthesis/deep-dives subdirectories. Fifteen comparison pages, thirteen language surveys, and seven concept syntheses were already present. Nothing was duplicated.
What was added
Six new files in raw/research-runs/:
empirical-validation-agent-language.pplx.md— the deep run that answers prompt 1 of prompts-mo-parallel-tracks (Perplexity session4a4e7abb, 520 lines). How to prove Mo beats a mainstream baseline with the same checks bolted on. Berger et al. reproduction protocol, cold-start experiment, maintainability experiment, human-review experiment.ecosystem-stdlib-platform-depth.pplx.md— the deep run that answers prompt 2 (Perplexity session45e714aa, 551 lines). Go/Rust/Elixir stdlib history, first-party kits (Laravel, Phoenix, shadcn), the recipes model, supply-chain defaults.agent-authoring-research-frontier.pplx.md— the deep run that answers prompt 3 (Perplexity session256a997f, 800 lines). Training-data gravity (79% Python, 9% no-resource pass@1), the language-for-LLMs thesis, structured diagnostics, verification-in-the-loop, effect systems and capabilities for agents.mo-parallel-tracks-brief-1-empirical.pplx.md,-brief-2-ecosystem.pplx.md,-brief-3-agent-authoring.pplx.md— the three short briefs Claude wrote for those deep runs, from Perplexity session2c696217.
One new page in research/prompts/:
- prompts-mo-parallel-tracks — pairs each brief with its deep run and lists the next-natural concept pages Fable might write.
Three adjacent Perplexity runs from the same week landed in raw/articles/, marked pplx-*, because they are context for Robert’s other projects rather than Mo design inputs: pplx-ai-agent-search-toolbox-2026-09.md (which retrieval APIs to use for AI agents), pplx-agent-sandbox-exe-dev-fly-sprites-2026-09.md (Exe.dev vs Fly.io Sprites for coding-agent sandboxes), and pplx-rust-vs-python-etl-data-engineering-2026-09.md (Rust for ETL and data engineering). They are here so the sha256 record is complete, not because they belong in the design.
What still needs doing
The three deep runs are inputs, not decisions. They read best in this order:
- agent-authoring-research-frontier first (the widest empirical picture; Giagnorio et al., Dan Luu, the Lingering Authority capabilities paper). Fold the new numbers into research-summary-2026-09 and case-against-new-languages.
- empirical-validation-agent-language second. Turn it into a concept page — provisional slug
empirical-validation-plan— that names the null-hypothesis, cold-start, maintainability, and human-review experiments, with baselines and pre-registered thresholds. Update d28 with named experiments; it is currently a slogan. - ecosystem-stdlib-platform-depth third. Turn it into a concept page — provisional slug
ecosystem-strategy— that answers Q11 as a stdlib scope and a kit list, cross-linked to d34 and d35.
None of this touches the decision-log until Fable ratifies it in a proper session. This one is the ingestion, not the read.