A true story · Aug 2025—Jul 2026

THINK
RUN.

What does it take to make the web runnable by intelligence? 323 days of building, breaking, reviewing, and remembering.

Enter the build

It started with a stubborn idea: the web should not merely be read. It should be operable — by people, by agents, from the cloud or the machine in front of you.

0000
Verified commitsNot a weekend project.
01 / THE SIX ACTS

Momentum has
a shape.

ThinkRun did not arrive in one flash. It accumulated capability — service, resilience, local control, sharing, recording, distribution — each layer making the next one possible.

01GenesisAug 20—Nov 30, 2025Browser automation service
02Core EngineDec 1—Jan 31Auth · resilience · video
03Extension & LocalFeb 1—Mar 31Chrome · bridge · Loom sharing
04Sharing & MCPApr 1—May 15oEmbed · MCP parity · upload
05Recording PipelineMay 16—Jun 17Web recording · Activity Hub
06DistributionJun 18—Jul 9SEO · connectors · pricing

The machine behind
the machine.

Building an agent-ready browser was itself an agent-scale collaboration. Provider-recorded usage reveals the otherwise invisible material of the work.

16.84Brecorded development tokens
Recorded sessions2,161
CoverageJan—Jul ’26
Peak commit week155
Merged PRs572
GPT-5.2 GPT-5.3-CODEX GPT-5.4 GPT-5.4-MINI GPT-5.5 CLAUDE FABLE 5 CLAUDE OPUS 4.8 CLAUDE SONNET 4.6 CLAUDE SONNET 5  
02 / THE MOST IMPORTANT FEATURE
Not speed.
Not scale.
The memory to know
when you’re wrong.

The project drew from 264 memory pages: threads, daily logs, architecture decisions, failures, corrections. Git recorded what changed. Memory preserved why.

A later audit found numbers that looked plausible but were not verified. They were corrected. The lesson became part of the product: trust the claim verify the artifact.

323 days later · still running

Make the web
move.

Enter ThinkRun Every number on this page comes from the project record · Snapshot July 2026