A Valid Tool Call Can Still Be an Invalid Effect
Permission to call a tool is only the beginning. What must be checked before a consequential action is allowed to take effect?
Read at ACM QueueFOUNDER, HALTSEAL
I’m Scott Lee. I build systems for the moment software acts in the real world. Today, that means HALTSEAL: execution control before money moves.
01 CURRENT WORK
A new route is not
a new right to pay.
A $10,000 payout times out. The first provider cannot yet confirm whether it paid. A second path is available. Sending again could pay the same obligation twice.
I built HALTSEAL to put a decision at that boundary. Across the paths it governs, it checks the exact payment against current authority and evidence about earlier attempts before release.
May this exact payment
proceed now?
Decision before action.
Proof before consequence.
Payments are the starting point. My ambition is a dependable execution layer for AI: a clear right to act, checked while the action can still be stopped.
Where the work standsPublic demo and synthetic evidence.
Customer-hosted validation is the next milestone.
02 WRITING & RESEARCH
The ideas behind
the systems I build.
Permission to call a tool is only the beginning. What must be checked before a consequential action is allowed to take effect?
Read at ACM QueueIf AI gives us back an hour, I hope we leave some of it open.
Why I started with payments, and what dependable delegation requires.
My individual Internet-Draft submitted to the IETF proposes verifiable authorization evidence before AI-agent and workload effects are committed.
Review the public evaluation codeMy related work includes 25+ U.S. patents issued or allowed for issuance, spanning machine authorization and execution controls.
Essays on how permission, incentives and the final payment can diverge.
03 THE FOUNDER
The person on the receiving end
is always part of the system.
I build around a simple belief: powerful systems should give people more agency, and earn the trust they ask for.
I founded Gooroo to make good teaching easier to reach. We built a marketplace that reached more than $1 million in annual recurring revenue and a network of more than 1,000 tutors.
Building and operating that business taught me that trust lives in the details: what was promised, what was delivered and whether the person on the other side could rely on it.
That experience shapes HALTSEAL. I wrote the core system myself. My related inventions and public technical work explore the same question: what gives a machine the right to act? The ambition is large; the work starts with one payment and a decision we can inspect.
I studied Operations Research at Columbia Engineering, worked in treasury at J.P. Morgan and served as a sergeant in the Republic of Korea Army. Those experiences shape how I build: think clearly, follow through and stay accountable to the people affected.
04 LET’S TALK
I’m looking for people who want to make this real: payment operators, builders and investors who care about the infrastructure behind trustworthy AI. If that’s you, I’d like to hear from you.
Have a payment workflow to discuss?