SALVO
What I BuildHow I Can Helpvs ChatGPTCasesResearchAbout
Contact⌖
SALVO
What I BuildHow I Can Helpvs ChatGPTCasesResearchAboutContact

The Record · verified benchmarks

Small models. Local hardware. Measured against frontier systems.

Every number reproducible on machines I own — and the failed approaches are reported right next to the results. Legal reasoning, on-chain execution, poker, and canine pose — each head-to-head against the frontier system it beats.

See the record →

Essay 01

Robots Aren't Ten Years Away

Perception and reasoning closed the robotics gap — deployment will be task-specific machines, not metal people.

Read essay →

Essay 02

AI Is an Attractor, Not a Multiplier

Productivity compression, homogenization, and slowdown are three faces of one mechanism.

Read essay →

Essay 03

The Say/Do Gap of Frontier Language Models Is Model-Specific, Not Yet Model-Intrinsic

What 27,750 adversarial reasoning traces reveal about agent faithfulness, theory of mind, and calibration.

Read essay →

Essay 04

The Expansion Regime

A field report from the tail the productivity literature never sampled — scope that would not otherwise exist.

Read essay →

Essay 05

Why Most AI Training Programs Fail

A cognitive mismatch hypothesis for why enterprise AI training produces 2–3× gains instead of 10–20×.

Read essay →

SALVO

I build measured AI systems for owner-operated businesses.

© 2026 WORKING SOFTWARE, NOT RECOMMENDATIONS

Hiring
jake@salvo.systems