Jason Bullen

Building FirstHelm in public: an AI agent control plane for governing autonomous AI. Real numbers, honest lessons, including the failures.

[Jason Bullen @FirstHelm_owner]

@iamdecatalyst "The one time I didn't" is the whole governance problem in one sentence. The setup that survives: agent opens PRs freely, but merges touching auth require a human yes — so skipping a day never costs you the auth stack.

[Jason Bullen @FirstHelm_owner]

@andilosing Every agentic dev has a version of this. The scary part isn't the deleted dev DB — it's "it also had access to production." Backups are the recovery plan. Blast radius is the prevention plan: dev-only creds for dev agents, human yes before anything destructive.

[Jason Bullen @FirstHelm_owner]

@samueldeploys Fair point. The pattern I keep seeing is that governance gets bought the week after an agent does something it shouldn't, not before. The teams posting publicly about their incident are already feeling that pain - and they usually haven't picked a tool yet. That's my GTM homework

[Jason Bullen @FirstHelm_owner]

Day one on X. I'm Jason, building FirstHelm, an AI agent control plane: live monitoring, guardrails, human approvals for autonomous agents. Current status: product live, zero users, three weeks of lessons. I'll post the real numbers here, including the ugly ones.