Probabilistic Systems Engineering Start here

Orientation

Start Here

Gregory Tomlinson

Probabilistic Systems Engineering is a bounded body of work on AI-assisted software change: what explicit source-of-truth artifacts help preserve, where authority actually lives, and where contracts stop.

Start with the papers for the bounded research claim. Authority essays remain important, but they are not the first mechanism/proof entry.

Choose your path

Engineers / researchers

Start with the boundary paper, then move into the engineering and stability papers.

Contract Authority Under AI v0.7.3 Contract-Centered Iterative Stability v4.7.3 Contract-Centered Engineering v2.17 Authority, Execution, and Refusal

Managers / executives

Use the non-engineering bridge first, then return to the proof summary only if you want the bounded research claim.

What AI Gets Wrong When You Iterate v0.2 What is proven / what is not

Curious outsiders

Read the proof boundary first, then the boundary paper. Move into authority only if you want the broader conceptual framing.

Start with the proof boundary Contract Authority Under AI v0.7.3 Then read authority

Core reading path

  1. Contract Authority Under AI v0.7.3

    States the narrower result, the surviving claim, and the authority boundary that did not hold.

  2. Contract-Centered Iterative Stability v4.7.3

    Shows the repeated drift mechanism and the boundary where prompts stop carrying invariant scope.

  3. Contract-Centered Engineering v2.17

    Explains why explicit source-of-truth artifacts still matter even when they are not the final authority surface.

  4. Authority, Execution, and Refusal

    Moves from the bounded engineering result into the broader authority framing.

Contracts are support artifacts. Replication is rerun support. Archive browsing is optional unless you care about lineage or earlier claim framing.

Key terms

These are the only terms you need up front. The full glossary remains in the replication material for portability and rerun support.

Authority
The place where a system can actually permit, refuse, or constrain action.
Contract
An explicit source of truth for what must stay true across change.
Invariant
A rule that must continue to hold after a change.
Decision surface
The place where acceptance, persistence, execution, or refusal actually occurs.
Replication
Portable rerun material for checking or transporting the work.
Proof-of-mechanism
Evidence that a repeatable failure mechanism exists, not a claim of universal statistical closure.