[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6},"actionData","errors","routes/public/w-agenda",{"_7":8,"_9":10,"_11":-5,"_12":13,"_14":15,"_16":17,"_18":19,"_20":21,"_22":23,"_24":25,"_26":27,"_28":8,"_29":8},"published",true,"event",{"_274":349,"_350":351,"_352":353,"_61":354,"_63":355,"_356":357,"_358":359},"accent","days",[336,337,338],"day","2026-09-15","dayLongLabel","Tuesday, September 15","window",{"_332":265,"_333":165,"_334":335},"rooms",[60,116,137],"blocks",[276,277,278,279,280,281,282,283,284,285,286,287,288],"legend",[271,272,273],"daySessions",[30,31,32,33,34,35,36,37,38,39,40,41,42],"scheduledAny","anyPublicSessions",{"_43":261,"_45":262,"_47":263,"_49":264,"_51":152,"_53":153,"_55":135,"_57":136,"_59":116,"_61":265,"_63":266,"_65":267},{"_43":245,"_45":246,"_47":247,"_49":248,"_51":52,"_53":54,"_55":249,"_57":250,"_59":116,"_61":251,"_63":215,"_65":252},{"_43":232,"_45":233,"_47":234,"_49":235,"_51":52,"_53":54,"_55":135,"_57":136,"_59":60,"_61":215,"_63":216,"_65":236},{"_43":226,"_45":227,"_47":228,"_49":229,"_51":152,"_53":153,"_55":135,"_57":136,"_59":116,"_61":215,"_63":216,"_65":230},{"_43":211,"_45":212,"_47":213,"_49":214,"_51":133,"_53":134,"_55":135,"_57":136,"_59":137,"_61":215,"_63":216,"_65":217},{"_43":198,"_45":199,"_47":200,"_49":201,"_51":52,"_53":54,"_55":135,"_57":136,"_59":60,"_61":181,"_63":182,"_65":202},{"_43":185,"_45":186,"_47":187,"_49":188,"_51":152,"_53":153,"_55":135,"_57":136,"_59":116,"_61":181,"_63":182,"_65":189},{"_43":177,"_45":178,"_47":179,"_49":180,"_51":133,"_53":134,"_55":135,"_57":136,"_59":137,"_61":181,"_63":182,"_65":183},{"_43":164,"_45":165,"_47":166,"_49":167,"_51":52,"_53":54,"_55":135,"_57":136,"_59":60,"_61":138,"_63":118,"_65":168},{"_43":148,"_45":149,"_47":150,"_49":151,"_51":152,"_53":153,"_55":114,"_57":115,"_59":116,"_61":138,"_63":154,"_65":155},{"_43":129,"_45":130,"_47":131,"_49":132,"_51":133,"_53":134,"_55":135,"_57":136,"_59":137,"_61":138,"_63":118,"_65":139},{"_43":110,"_45":111,"_47":112,"_49":113,"_51":52,"_53":54,"_55":114,"_57":115,"_59":116,"_61":117,"_63":118,"_65":119},{"_43":44,"_45":46,"_47":48,"_49":50,"_51":52,"_53":54,"_55":56,"_57":58,"_59":60,"_61":62,"_63":64,"_65":66},"id","demo-s-36","friendlyNum",37,"title","Panel: What Breaks When AI Products Meet Real Users","descriptionHtml","

Four practitioners on the gap between the demo and the deployment: the failures that only appear at scale, the support burden nobody budgets for, and what they would build differently. Audience questions from the start.

","trackId","demo-track-0","trackName","AI Engineering","formatId","demo-format-4","formatName","Panel (45 min)","roomName","Main Stage","startsAt",1789510500000,"endsAt",1789513200000,"speakers",[67,68,69,70],{"_71":102,"_73":103,"_75":104,"_47":105,"_78":106,"_80":107,"_82":108,"_84":109,"_86":8},{"_71":95,"_73":96,"_75":97,"_47":98,"_78":99,"_80":100,"_82":101,"_84":85,"_86":87},{"_71":88,"_73":89,"_75":90,"_47":91,"_78":92,"_80":93,"_82":94,"_84":85,"_86":87},{"_71":72,"_73":74,"_75":76,"_47":77,"_78":79,"_80":81,"_82":83,"_84":85,"_86":87},"personId","demo-p-20","firstName","Aisha","lastName","Rahman","VP Engineering","company","Loomis","bio","Aisha grew Loomis engineering from twelve people to ninety while shipping the company's first AI product. She writes mostly about the parts of that which went badly.","photoFileId","demo-photo-20","role","panelist","isPrimary",false,"demo-p-13","Kwame","Mensah","Director of Engineering","Northwind Health","Kwame leads the clinical AI group at Northwind, where every model output is reviewed by a clinician before it reaches a chart. He is candid about what regulation does and does not actually slow down.","demo-photo-13","demo-p-5","Tom","Achebe","CTO","Forgeline","Tom co-founded Forgeline after leading platform engineering at two payments companies. He is the reason Forgeline had an evaluation suite before it had a marketing site.","demo-photo-5","demo-p-10","Grace","Liu","AI Product Lead","Meridian","Grace has shipped AI features to twelve million users and killed about as many that never earned their place. She writes about interface design for systems that answer differently every time.","demo-photo-10","moderator","demo-s-24",25,"Reading Eval Traces Like a Detective","

A short, practical walkthrough of how to read a failing evaluation trace: what to look at first, which signals lie, and how to turn one bad output into a reproducible test case.

","demo-format-2","Lightning Talk (10 min)","Room 2A",1789504500000,1789505100000,[120],{"_71":121,"_73":122,"_75":123,"_47":124,"_78":125,"_80":126,"_82":127,"_84":128,"_86":8},"demo-p-4","Elena","Vasquez","ML Engineer","Braid Research","Elena works on retrieval and evaluation at Braid. She published an open benchmark suite after getting tired of numbers that never reproduced on her own hardware.","demo-photo-4","speaker","demo-s-6",7,"Developer Platforms People Actually Adopt: Lessons From Three Failures","

We built three internal platforms nobody used before one that stuck. The difference was never features — it was migration cost, escape hatches, and letting teams keep their own deploy buttons.

","demo-track-2","Developer Experience","demo-format-1","Talk (30 min)","Room 2B",1789503300000,[140],{"_71":141,"_73":142,"_75":143,"_47":144,"_78":145,"_80":146,"_82":147,"_84":128,"_86":8},"demo-p-6","Sofia","Lindqvist","Engineering Manager","Norrsken Data","Sofia manages the developer-experience group at Norrsken, a team she started by asking every engineer which internal tool they would delete first. She has been an SRE, a compiler engineer and, briefly, a professional oboist.","demo-photo-6","demo-s-23",24,"Ten Wrong Turns Migrating a Data Platform","

A fast tour of the ten decisions we would make differently, from partition keys to the migration window we scheduled during a holiday. Ten minutes, ten mistakes, no filler.

","demo-track-1","Platform & Infra",1789503900000,[156],{"_71":157,"_73":158,"_75":159,"_47":160,"_78":161,"_80":162,"_82":163,"_84":128,"_86":8},"demo-p-15","Diego","Ramirez","Data Platform Engineer","Verdant","Diego moved Verdant's analytics stack off a warehouse it had outgrown and wrote down every wrong turn on the way. He mentors at two Bay Area programs.","demo-photo-15","demo-s-8",9,"Prompt Injection Is a Solved Problem (If You Do These Five Things)","

Provocative title, sober content: capability scoping, output mediation, and the three-layer defence we run in production. Includes live demos of attacks that still work against naive agents in 2026.

",[169],{"_71":170,"_73":171,"_75":172,"_47":173,"_78":174,"_80":175,"_82":176,"_84":128,"_86":8},"demo-p-8","Amara","Diallo","Security Researcher","Halcyon Labs","Amara leads red-team work on agent systems at Halcyon. Her disclosure of the tool-confusion class of prompt injections is why several vendors now scope tool permissions per call.","demo-photo-8","demo-s-10",11,"Design Systems for AI Products: Calm Interfaces for Chaotic Models","

Non-determinism does not excuse jittery interfaces. Tokens, motion budgets and content design for products where the backend is a probability distribution and the user still needs to feel in control.

",1789496100000,1789497900000,[184],{"_71":102,"_73":103,"_75":104,"_47":105,"_78":106,"_80":107,"_82":108,"_84":128,"_86":8},"demo-s-1",2,"Taming 40-Minute CI: Incremental Builds at Monorepo Scale","

Our monorepo CI hit forty minutes and engineers stopped trusting it. We rebuilt it around content-addressed caching and got to six. A decision framework for which incremental-build investments actually pay off, and which ones are cargo cult.

",[190],{"_71":191,"_73":192,"_75":193,"_47":194,"_78":195,"_80":196,"_82":197,"_84":128,"_86":8},"demo-p-1","Marcus","Okafor","Staff Developer Advocate","Cloudreach Labs","Marcus spent six years on build infrastructure before moving into advocacy, and still maintains the caching tool that came out of it. Most of his talks open with a CI pipeline somebody let rot for two years.","demo-photo-1","demo-s-3",4,"Multi-Agent Systems Are Just Distributed Systems (And That's Good News)","

Everything your team already knows about queues, retries, idempotency and backpressure applies directly to agent orchestration. A tour of production multi-agent failures and the boring distributed-systems fixes that solved each one.

",[203],{"_71":204,"_73":205,"_75":206,"_47":207,"_78":208,"_80":209,"_82":210,"_84":128,"_86":8},"demo-p-3","Yuki","Tanaka","Head of Platform","Kitsune Cloud","Yuki built Kitsune's multi-region control plane and now leads the team that keeps it boring. Before that: eight years of distributed databases and one very educational New Year's Eve outage.","demo-photo-3","demo-s-7",8,"Streaming Structured Output: UX Patterns Beyond the Token Firehose","

Users do not want to watch JSON assemble itself. Patterns for progressive disclosure, optimistic rendering, and graceful repair when the model changes its mind halfway through a response.

",1789493400000,1789495200000,[218],{"_71":219,"_73":220,"_75":221,"_47":222,"_78":223,"_80":224,"_82":225,"_84":128,"_86":8},"demo-p-7","Raj","Patel","Founding Engineer","Tandem Agents","Raj was employee number two at Tandem and wrote the streaming layer the product is now known for. He has rewritten it three times and expects a fourth.","demo-photo-7","demo-s-4",5,"The Edge Database You Already Have: SQLite Patterns for Global Apps","

Read replicas at the edge, single-writer semantics, and what happens when you stop pretending you need a cluster. Real latency numbers from three production migrations, including the one we rolled back.

",[231],{"_71":121,"_73":122,"_75":123,"_47":124,"_78":125,"_80":126,"_82":127,"_84":128,"_86":8},"demo-s-2",3,"Your AI Pair Programmer Is Lying to You: Verification Patterns That Scale","

Generated code compiles, passes review, and still breaks production. We catalogued a year of agent-written incidents and built verification layers that catch what human review misses — without slowing anyone down.

",[237],{"_71":238,"_73":239,"_75":240,"_47":241,"_78":242,"_80":243,"_82":244,"_84":128,"_86":8},"demo-p-2","Dana","Kowalski","Principal Engineer","Substrate AI","Dana runs the code-generation platform at Substrate, where about a third of merged pull requests are now machine-authored. Her work is mostly the verification layer that makes that survivable.","demo-photo-2","demo-s-0",1,"Context Engineering in Production: What Actually Survives Contact With Users","

A year of shipping agent products taught us that context design outlives model choice. This talk covers the patterns that survived: compaction budgets, retrieval boundaries, and the observability you need before you can tune anything.

","demo-format-0","Keynote (45 min)",1789490700000,[253],{"_71":254,"_73":255,"_75":256,"_47":257,"_78":258,"_80":259,"_82":260,"_84":128,"_86":8},"demo-p-0","Priya","Raman","Staff Engineer","Latticework Systems","Priya leads the agent platform team at Latticework, where she rebuilt the company's retrieval stack around explicit context budgets after a rewrite that took two attempts. She writes the Context Notes newsletter and has been shipping LLM products since the first public API keys.","demo-photo-0","ae62f91f-f263-4348-a833-bf6065aa6977",51,"UPDATED: Taming 40-Minute CI: Incremental Builds at Monorepo Scale","

Our monorepo CI took 40 minutes on a good day. This talk walks through how we cut it to 6 minutes with content-addressed caching, remote execution, and a test-selection model — including the two migrations that failed first. You'll leave with a decision framework for which incremental-build investments pay off at which repo sizes, and the graphs to convince your platform team. Updated: now includes 2026 benchmark data. This session now includes a live demo of remote build caching.

",1789488000000,1789489800000,[268],{"_71":269,"_73":255,"_75":256,"_47":-5,"_78":-5,"_80":270,"_82":-5,"_84":128,"_86":8},"a0171bb2-3c93-44c0-9d44-f8265a64bba4","Priya Raman is a Principal Engineer at Latticework Systems where she leads the build-tooling platform team. She previously maintained the open-source task runner 'gantry' and has spoken at over a dozen developer conferences on build systems, CI reliability, and developer productivity metrics.",{"_43":52,"_274":54,"_275":246},{"_43":152,"_274":153,"_275":186},{"_43":133,"_274":134,"_275":233},"name","tone",{"_43":232,"_289":60,"_47":234,"_65":331,"_291":308,"_293":309,"_275":246,"_53":54,"_295":87,"_296":310,"_298":299,"_300":301,"_302":246},{"_43":198,"_289":60,"_47":200,"_65":330,"_291":304,"_293":305,"_275":246,"_53":54,"_295":87,"_296":306,"_298":299,"_300":301,"_302":246},{"_43":164,"_289":60,"_47":166,"_65":329,"_291":292,"_293":294,"_275":246,"_53":54,"_295":87,"_296":297,"_298":299,"_300":301,"_302":246},{"_43":44,"_289":60,"_47":48,"_65":325,"_291":326,"_293":327,"_275":246,"_53":54,"_295":87,"_296":328,"_298":322,"_300":301,"_302":246},{"_43":261,"_289":116,"_47":263,"_65":319,"_291":323,"_293":324,"_275":186,"_53":153,"_295":87,"_296":301,"_298":299,"_300":301,"_302":246},{"_43":245,"_289":116,"_47":247,"_65":319,"_291":320,"_293":321,"_275":246,"_53":54,"_295":87,"_296":322,"_298":322,"_300":301,"_302":246},{"_43":226,"_289":116,"_47":228,"_65":311,"_291":308,"_293":309,"_275":186,"_53":153,"_295":87,"_296":310,"_298":299,"_300":301,"_302":246},{"_43":185,"_289":116,"_47":187,"_65":318,"_291":304,"_293":305,"_275":186,"_53":153,"_295":87,"_296":306,"_298":299,"_300":301,"_302":246},{"_43":148,"_289":116,"_47":150,"_65":316,"_291":317,"_293":294,"_275":186,"_53":153,"_295":87,"_296":297,"_298":315,"_300":301,"_302":246},{"_43":110,"_289":116,"_47":112,"_65":311,"_291":312,"_293":313,"_275":246,"_53":54,"_295":87,"_296":314,"_298":315,"_300":301,"_302":246},{"_43":211,"_289":137,"_47":213,"_65":307,"_291":308,"_293":309,"_275":233,"_53":134,"_295":87,"_296":310,"_298":299,"_300":301,"_302":246},{"_43":177,"_289":137,"_47":179,"_65":303,"_291":304,"_293":305,"_275":233,"_53":134,"_295":87,"_296":306,"_298":299,"_300":301,"_302":246},{"_43":129,"_289":137,"_47":131,"_65":290,"_291":292,"_293":294,"_275":233,"_53":134,"_295":87,"_296":297,"_298":299,"_300":301,"_302":246},"room","Sofia Lindqvist","timeLabel","1:15 PM – 1:45 PM","startLabel","1:15 PM","conflicted","startMin",255,"minutes",30,"lane",0,"lanes","Grace Liu","11:15 AM – 11:45 AM","11:15 AM",135,"Raj Patel","10:30 AM – 11:00 AM","10:30 AM",90,"Elena Vasquez","1:35 PM – 1:45 PM","1:35 PM",275,10,"Diego Ramirez","1:15 PM – 1:25 PM","Marcus Okafor","Priya Raman","9:45 AM – 10:30 AM","9:45 AM",45,"9:00 AM – 9:30 AM","9:00 AM","Grace Liu, Tom Achebe, Kwame Mensah, Aisha Rahman","3:15 PM – 4:00 PM","3:15 PM",375,"Amara Diallo","Yuki Tanaka","Dana Kowalski","startMs","startHour","endHour",16,{"_339":15,"_341":348,"_343":17},{"_339":345,"_341":346,"_343":347},{"_339":340,"_341":342,"_343":344},"key","2026-09-17","label","Thu, Sep 17","longLabel","Thursday, September 17","2026-09-16","Wed, Sep 16","Wednesday, September 16","Tue, Sep 15","AI Builders Summit 2026","slug","ai-builders-summit-2026","location","Fort Mason Center, San Francisco",1789473600000,1789646400000,"timezone","America/Los_Angeles","notice","Demo conference. The talks, speakers and reviews below are invented for demonstration — nobody named here submitted anything."]