Transcript — AIOS Pod Model (Mansel Scheffel)
architecture client tooling api
R&D Learnings
Institutional knowledge captured from R&D evaluations — surprises, reusable patterns, mistakes, and discovered constraints.
8 learnings | unknown (8)
R&D Learning Card — AI Operating System (Pod Model)
architecture framework methodology client
Auto-Research Pattern Requires Upfront Eval Criteria
The auto-research pattern is powerful but only when paired with explicit, binary evaluation criteria defined before the first iteration. "Make it better" loops diverge; "does it pass these 10 checks?"...
evaluation methodology tooling pattern
Product Naming Carries Political Weight — "Employee" vs "Talent"
Product naming in AI is politically loaded in ways that pure technology evaluation misses. Terms that are technically accurate ("employee" — it does the work of an employee) can be commercially toxic ...
evaluation client naming process
HTML Deliverables Dramatically Outperform Markdown for Client Reception
Format is not just presentation — it's perceived competence. For consulting deliverables, interactive HTML with professional styling signals "we built something for you" while markdown signals "we wro...
evaluation documentation client delivery
LeanIX v4 Official Docs Lag Behind API Reality
For enterprise SaaS platforms undergoing major version changes (v3→v4), official documentation trails the actual API by weeks or months. The reliable source of truth is the API itself — not PDFs, not ...
evaluation framework documentation api
Learning — Plugin Bundle Delivery Constraints
- **`claude plugin validate` is the cheap, authoritative feedback loop.** A schema-invalid manifest fails *silently* at runtime — never ship a bundle without the validator passing. It is now a hard ga...
delivery
Shipping Before Optimizing Unlocks Revenue; R&D Delays Don't
For a factory-stage startup, the bottleneck is almost never product quality — it's delivery velocity. Every day an R&D exploration delays a client handover is a day of blocked revenue with zero return...
client pattern process delivery
No learnings match your search.
About R&D Learnings
Learnings are documented by Riley (R&D Analyst) at Stage 5 of the R&D pipeline when an evaluation outcome reveals something worth preserving as institutional knowledge. Not every evaluation produces a learning — only when:
- 1. The outcome surprised us — result differed from the prediction
- 2. A reusable pattern emerged — applicable beyond the specific case
- 3. A mistake was made — wrong decision or flawed scoring
- 4. A new constraint was discovered — technical, legal, or organizational