Eric Allam

we're opening up more roles, hiring remote or in person in London 🇬🇧, building some very cool open source stuff and working on hard/medium/easy problems (we have them all). DM me and please share 🙏

Trigger.dev lets developers add AI agents to their products with a simple SDK, handling execution, long-running workflows, and reliability so they don't have to. Over 90% of their usage now comes from agent workflows, and they recently announced their $16M Series A led by Standard Capital. In this episode of Founder Firesides, co-founders @mattaitken and @maverickdotdev sat down with YC's @koomen to talk about three versions of the product before finding product market fit, how building async infrastructure for two years accidentally put them in the perfect position for the agent era, and why they think the future of computing is programmatic checkpoint and restore — freezing and resuming compute on demand.

Bolto is a single platform that provides AI-native recruiting, HR and payroll in the same product, from YC S23. I just posted an interview with @mrinalsingh02 and @jakeatbolto, the founders of @boltohq, about their recent $ 12M Series A from @Standard_Cap. Link below

. @mattaitken from @triggerdotdev talking durable agents - @aiDotEngineer Europe

📝Detailed write up of how we ditched git worktrees in favor of @gitbutler and how we use it with agents here: trigger.dev/blog/parallel-…

We've decided to get into the 👟 business

We led the Series A for Insight Health. Insight Health builds AI agents for specialty care. The founders include a YC alum, the former head of cloud infrastructure at Twilio, and two practicing physicians. Welcome to @Standard_Cap, Insight Health!

@acadb3k @pugson @Zai_org @triggerdotdev we also do support idempotency keys when triggering tasks: trigger.dev/docs/idempoten…

@acadb3k @pugson @Zai_org @triggerdotdev Typically for that kind of thing you could use idempotent writes to the database (e.g. upserts) but I'm not sure what credit drift is

Vercel is great, but... My AI app generates 5-10 screens via @Zai_org Vercel's timeout kills it in prod. Need a service for long-running AI tasks: Supports Next.js timeout 10 min+ Railway? Render? VPS? What are you using? #saas #developers #nextjs

@acadb3k @pugson @Zai_org @triggerdotdev Use batchTriggerAndWait and do each screen in a separate task like this:

@maverickdotdev @pugson @Zai_org @triggerdotdev wow so glad to see you here! that’s exactly what I’m planning to do. thanks! I have another question, mind if I ask? for 6-8 screens (1-2 min each), do you recommend 1 run = 1 job (sequential) or 1 screen = 1 job (parallel)? Which is better for production?

@acadb3k @pugson @Zai_org @triggerdotdev Do you mean stream updates to convex as in like saving the llm stream chunks to a the database? You can definitely do that with us, its a full Node.js process so you can do pretty much anything

@pugson @Zai_org @triggerdotdev sequential ai generation takes 5-10 mins can i stream updates to Convex from a Trigger task in real time? and how does it handle 3k users without hitting LLM rate limits? Is this the best production safe path?

Been running AI automation on GitHub PRs - generating them, reviewing them. Was doing it in GitHub Actions. Debugging GitHub Actions is a slog. Rerunning a failed job is a hassle. Observability beyond pass and fail requires parsing logs. Switched to @triggerdotdev. Night and day. Easier to set up, integrate locally, rerun, and actually see what's happening. The DX is much better.

Switching from Node to @bunjavascript gave us a 5x speed boost (28x better max latency) but there was a gotcha… Pending response promises don't resolve on client disconnect. @nodejs handles this automatically. Fix: resolve disconnects with 499. trigger.dev/blog/firebun

@NathanFlurry @triggerdotdev Ah man someone beat us to it! We need to publish a demo

Another cool use case we didn't consider: company started using Secure Exec with @triggerdotdev 👀

🤘 Rivet Secure Exec (V8 isolates): Securely execute Node.js anywhere (Railway, Vercel, Hetzner, Kubernetes) vs 🔥 Firecracker (microVMs): Full Linux operating system, but requires specialized infrastructure

@bmcnew_ @triggerdotdev How about us?

Ran an agent audit on @triggerdotdev

Just launched: our Vercel integration for Trigger. Push code. @vercel deploys your app. @triggerdotdev deploys your tasks. Env vars sync both ways. Your app never goes live with mismatched task versions. ↓

@nicoalbanese10 Have something for this coming soon 👀

Building durable agents with resumable streams, aborting, and persistence is hard

Do you think Claude gives us ratings on how well we are doing this session?

SF is for coding agents, London is for enterprise agents

@wishee0 @boristane he would but unfortunately he's going to use vue

@boristane you better have one heck of a good ui at nominal!!

I need this damn