Synthetic Intuition Core: MCP Workflow Pair
by DeviousDevv303
A falsifiable conviction system for AI agents — provenance-gated evidence, calibration bounds, and a disproof court — installable as two MCP workflow servers.
I am an automation developer.
by DeviousDevv303
A falsifiable conviction system for AI agents — provenance-gated evidence, calibration bounds, and a disproof court — installable as two MCP workflow servers.
by DeviousDevv303
Your agent records predictions before outcomes and lets its graded track record set its confidence — no more "90% sure" on a 1-of-2 record.
by DeviousDevv303
Teach your agent to classify a failure before reacting to it — broken method or disproven claim — so it stops quitting good work and defending dead work.
by DeviousDevv303
Plan and carry out dependency upgrades without breaking the build. Every breaking-change claim is tied to a source, every affected call site is found in your code, and each step has a check and a way back. Use for major version bumps, CVE patches, framework migrations, and clearing out stale package
by DeviousDevv303
Debug a failure to a cause that is demonstrated, not guessed. Reproduce it, narrow it with experiments that could prove you wrong, fix the cause instead of the symptom, and leave a regression test. Use for failing tests, crashes, wrong output, flaky behavior, and 'it worked yesterday'.
by DeviousDevv303
Turns a spec, function, or diff into a test suite engineered to fail against near-miss (mutated) implementations, not just obvious breakage. Use when code needs tests that prove behavior rather than exercise it.
by DeviousDevv303
Review code changes (a diff, a branch, files, or pasted code) for real defects. Every finding carries a line citation and a concrete failing scenario; anything unproven is labeled unverified. Use for pre-merge review, 'review my changes', bug hunts, and security passes.
by DeviousDevv303
Bootstrap Monte Carlo for strategy evaluation — distributions and stated artifacts, not one lucky equity curve.
by DeviousDevv303
The reasoning layer proposes; the wrapper disposes. Mechanical floors for agents that can spend money.
by DeviousDevv303
Turn every decision — human or AI — into a calibrating dataset. Gradeable in advance, or it doesn't count.
by DeviousDevv303
Orchestrate the swarm, journal the decisions, simulate the policies, floor the autonomy — one verified arc.
by DeviousDevv303
Multi-agent teams that verify everything against the source — zero "trust me" handoffs.
by DeviousDevv303
Do the same agent work for a fraction of the tokens — seven habits, one counted worked example, no invented savings.