Two builders, one obsession: less waste.
Tidbit reports Claude savings by measured workload. Tracked runs cover one large tool-heavy request and repeated stable context. Here’s why we built it, and who’s behind it.
fig. 01 - built, then bitten
Why we built Tidbit
AI has become core infrastructure, and its cost deserves evidence rather than a broad promise. In tracked runs, one tool-heavy request used fewer billed input tokens, while repeated stable context reduced cumulative billed cost.
One of us has spent a career cutting cloud bills; the other, building products people love. We built Tidbit to report outcomes by workload. In the cited lossless-profile runs, recorded fidelity checks passed; those checks are limited to those source runs and are not a general quality or latency guarantee.
I’m a builder at heart, happiest on greenfield work. I built my first websites on GeoCities in the ’90s, and 20+ years later I’m a father of two, now in greybeard territory, working across engineering, infrastructure, and product. AI has been my daily driver for two years, the best and most creatively free stretch of my career. Cost-conscious to a fault, I once cut a company’s cloud bill by well over half in a year. Tidbit is that same instinct aimed at AI: measure the workload, cut waste, and publish the result.
Matt is a product and design leader who has spent over a decade shipping software people actually enjoy - co-founding a restaurant-technology company and launching dozens of apps across iOS, Android, Apple TV, Roku and the web along the way. Equal parts designer, product manager, and brand-builder, he makes powerful tooling feel effortless. At Tidbit he owns the product and the experience - the console, the onboarding, and everything you touch.
Want to reach out?