the machine that builds failproof
an engineering session on the setup that writes most of failproof: a spec in, coding agents in parallel, policies on our own agents, and traces that turn a failed run into a better spec.
most of the code in failproof is written by agents. a spec goes in, several coding agents work on it in parallel, and what comes out gets shipped. this is the whole loop, with real runs on screen — including the failed ones.
what it's about
most of the code in failproof is written by agents. a spec goes in, several coding agents work on it in parallel, and what comes out gets shipped.
nivedit jain, co-founder and cto, walks through the whole setup and how we run it day to day: the loop end to end, the policies we point at our own agents, and the traces that feed the next spec. this is an engineering session, not a product demo.
- 01the loop, end to end
how a piece of work goes from a spec to shipped code, and what a human still does at each handover.
- 02running agents in parallel without chaos
how several agents work on the same codebase at once, how they avoid stepping on each other, and what happens when two of them disagree.
- 03eating our own cooking
the policies we run on our own agents, what they block, what they correct, and the ones we had to rewrite because they got in the way.
- 04traces as the feedback loop
how failed runs from the machine turn into better specs and new policies, so the same mistake doesn't happen twice.
- 05what still breaks
the failure modes we haven't solved, the tasks we still don't hand to agents, and why.
- 06your questions, live
the last 20 minutes are open. bring your own setup and what you're stuck on.
how the day runs
all times pst · subject to change- 15 minthe machine, end to endtalk
how a piece of work goes from a spec to shipped code, where several agents run in parallel, and what a human still does at each handover.
- 25 minwalkthrough of real runs, including the failed onestalk
the policies we point at our own agents, how a failed run turns into a better spec, and the failure modes we have not solved.
- 20 minopen q&asocial
bring your own setup and what you're stuck on.
who's on
- HOSTSahar Morbond ai
questions
something else? ask on discord →do i need to use failproof to get anything out of this?
no. the session is about running coding agents at the scale a small team ships at, and failproof is the codebase we do it on.
the spec-to-ship loop, keeping parallel agents out of each other's way and feeding failed runs back into the next spec are the same problems on any stack.
what will i actually see?
real runs, including the failed ones, not a slide of a workflow diagram.
how work goes from spec to shipped, what our own policies block and correct, and the tasks we still do not hand to agents.
can i bring my own problem?
yes. the last 20 minutes are open, so bring your own setup and what you're stuck on.
the q&a starts at 6:40 pm pt.
how does registration work?
register on luma. every registration is subject to host approval.
it is online, so luma sends the joining link once you are in.
what is failproof ai?
failproof is your agent's oversight layer, steering it towards success. it traces your agents, finds failure patterns over time and sets up rules to prevent them from happening again.
the open-source cli is free: npm i -g failproofai.
MORE EVENTS
all events →- OCT06how we steer hermesengineering session · san francisco
- OCT20handling 10 million agent traces a dayengineering session · san francisco
- OCT24steering agents at runtimepaper discussion · bengaluru