Go-to-market operator. Builder. Skeptic of my own numbers.
I spent Receipt Verified 2026-10-098+ years in GTM, analytics and AI.PeriodSep 2018 to todayMethodCounted from the first role on the master resume.SourceMaster resume (Sep 2026) on the revenue side of software: program management and revenue operations at ServiceTitan, revenue operations at Simpplr, analytics at Vercel, and GTM strategy at HackerOne. The job was the same each time: turn noisy signals into decisions leaders could act on.
In January 2026 I started Mind Theory AI to build those systems myself. I write the specs and the acceptance tests, Claude Code does most of the typing, and every safety check has to fail on purpose before I trust it. That is how one person shipped a graph-intelligence platform, a legal search engine and a speech stack in a few months.
Seven rules, each with proof
- Make every check fail on purpose before trusting it.
“A guard you have only ever seen pass is indistinguishable from
true.”Every deletion gate was first pointed at a wrong destination to prove it would refuse.
- Verify, then act. Count the matches, not the "0 differences".
Receipt Verified 2026-10-09Deleted 1,044.9 GiB across 9,395 files, each only after a fresh per-file hash proof against the cloud copy.Period2026-09-13 to 2026-10-06MethodSum of 13 deletion-gate logs.SourceDeletion-gate logs deleted, each file only after a fresh hash proof; Receipt Verified 2026-10-09When DNS failed mid-job, the gate refused 459 deletes and made 0, then the job resumed and finished.Period2026-10-04MethodRefusal log of the affected run.SourceGate refusal log refused / deleted during a DNS outage.
- Don't let a model's best score come from making things up.
Whisper produced text on Receipt Verified 2026-10-09whisper-large-v3 produced text on 93.5% of 200 clips where the voice detector found no speech.Period2026-08-01Method200 VAD-negative clips from field recordings; insertion rate per model.SourceBenchmark logCaveat"No speech" means the VAD found none, so this is an upper bound. of no-speech clips, so an independent model now holds a veto.
- A delta without an interval is not a result.
Model swaps ship with paired bootstrap intervals: Receipt Verified 2026-10-09Moving translation to Gemma raised chrF++ on dialect sentences from 53.10 to 61.68.Period2026-09-28MethodDialect set, n = 35; paired bootstrap, p = 0.011. In-house set (n = 100): +1.6, not significant.SourceEvaluation log.
- Label the guess.
Ghostline flags every inferred link and shows the entity it came through.
- Measure the bottleneck before tuning.
Hidden quota errors went Receipt Verified 2026-10-09Found 403 quota errors visible only in verbose logs (210 in one job) and took them to 0 with a dedicated OAuth client.PeriodSep to Oct 2026MethodError counts in comparable upload logs.SourceUpload logs once measured; pipelines are sized to the slowest real link.
- Agent findings are reports, not facts.
I fact-checked Receipt Verified 2026-10-09Fact-checked 343 company claims before using any: 324 supported, 16 partly, 3 rejected.PeriodRun of 2026-09-15Method21 verification batches over 124 companies.SourceVerification outputs (JSON) company claims before using any of them. This site's receipts follow the same rule.
- GTM from the inside: pricing and packaging, forecasting, board reporting and launches at growth-stage software companies.
- Speed: full-stack products from spec to production with AI agents, with verification built in.
- Rigor: controls, receipts and honest negative results, in a market full of demos.
A partner who wants to build verifiable AI products in markets where trust is the purchase decision.
- Shipping Ghostline's graph features: semantic super-edges and an as-of time scrubber are next.
- Building Khun toward a packaged app.
- Open to AI product and GTM engineering leadership roles, co-founder conversations, and investor conversations.