Not what we believe. How we actually decide.

Philosophy explains why OSG builds the way it does. This is the mechanism underneath it — the pattern a new hire would infer simply by watching how decisions get made, checked, and revised.

01 — Principles

Six patterns, checkable against how OSG has actually operated.

Structure before content

A decision, a page, or a claim isn't finished when it sounds right — it's finished when its reasoning is defensible on its own. We'd rather delay something than publish it structurally sound but under-evidenced.

Defensibility wins every trade-off

When a more complete narrative and a more accurate one conflict, accuracy wins without exception. A correction is accepted immediately and fully, not softened or partially retained.

Evidence is binary, not weighted

A claim is either sourced or it isn't published. Plausibility is not a substitute for verification — something that sounds right is treated the same as something we don't yet know, until it's confirmed.

Uncertainty is named, not smoothed

An open gap is labeled and sequenced, not quietly filled with a reasonable-sounding guess. This applies to public content and internal planning alike — a milestone with no verified date says so rather than estimating one.

Products evolve by scrutiny, not increment

When a premise doesn't hold up, it gets repositioned, not patched. CapitalOS didn't accumulate features on top of a founder-tool concept — its entire framing changed once the original premise was tested and failed.

Sequencing over speed

Work ships in dependency order — one layer complete before the next begins — even when parallelizing would move faster. The cost of a half-finished layer is judged higher than the cost of moving slower.

Principle 05 isn't abstract — the CapitalOS thesis itself was repositioned twice under this exact discipline. See the Journal entry recording the second one.

02 — In practice

What consistently gets rejected, and what gets reinforced.

Rejected

Marketing register, in any form — hype adjectives, inflated claims, startup clichés.
Plausible-sounding but unverified detail, regardless of how minor or how well it would read.
Completing a narrative by inference when a stated fact is simply missing.

Reinforced

Explicit dependency mapping before execution — what has to be true before this can be built.
Source-labeling every claim by type: verified fact, stated account, observed pattern, or synthesis.
Revisiting and correcting earlier work in place, rather than letting an error persist alongside a later fix.

03 — Where this leads

This is the standard Documentation is written to.

These six principles are not aspirational. They are the actual basis for the Decision Framework published in Documentation, and for the structure every Journal entry follows: what changed, why, what alternatives existed, what decision was made, what evidence informed it, what happens next. Nothing downstream should contradict what's stated here — if it does, one of the two is wrong and needs to be fixed.

Read the actual documentation →