A 24-hour prescription check took months to build by hand
Engineers spent the time on queues and timers, not medicine, and that frustration led Tony to co-found Inngest.
The Peel with Turner Novak · The Hidden Layer Every AI Agent Runs OnTony, the co-founder of Inngest, previously led engineering at a healthcare business where every action left a trail. The company was highly event driven because that record provided a clear audit trail of care and decisions. Inngest sells developers a simpler way to run multistep processes, and it grew out of that earlier frustration.
When a patient completed an intake document, when a doctor approved one, or when a patient requested medication, each event set off a chain of conditional next steps. One rule sounded especially simple. If a doctor had not approved a prescription within 24 hours, follow up on that request.
A one-line rule hid months of work. Tony said that check "was literally one of the worst things to build." His team spent months getting those basic flows to work, not because the medicine was complicated, but because the software had to remember a request, watch the clock, and start the right follow up when nothing had happened.
That work meant rebuilding the same queue-based plumbing he had built before. Around 2019, Tony said, he had spent months rebuilding that infrastructure again and again. Queues are waiting lines that hold computing jobs until another part of the system can handle them, and the classic example, still widely used, is SQS.
The Inngest approach is built around step functions, a general way to build multistep processes that remember where they left off. In that design, an incoming event starts a program that can pause until something happens inside the product. It then wakes up on its own when the product reports back, or when time runs out, and the team can see that the approval never arrived. For that prescription check, Tony said the work "would have taken a day or two."
For Tony, the point was practical. The team would no longer have to hand-build the waiting, remembering, and resuming. The program could start when the prescription was requested, pause for the approval, and resume automatically when the approval was missing.
Questions people ask
What does Inngest actually sell?
Inngest provides developers with building blocks for running multistep processes that remember their progress, retry failed steps, and wait for later events. Developers write the steps in code, and Inngest handles the queues, timing, state, and recovery underneath.
Why does software need queues for something as simple as waiting 24 hours?
A computer may need to wait hours or days for a human approval, a payment confirmation, or another service to respond. A queue stores that unfinished work safely so it is not lost if a server restarts, and lets another part of the system pick it up later and continue.
What is SQS?
SQS stands for Simple Queue Service, a widely used message queue service from Amazon Web Services. Software sends it a message about work to do later, and other software pulls messages from it when ready, which makes it useful for spreading work across many machines.