Hot take: I think it's still important to understand the code that our agents write!
In this mega thread (based on my AIE talk today), I will explain why that's the case, and show some ideas for how to efficiently understand code. Alright, let's dive in. 1/
The team absolutely cooked on this one!
I just built a web-agent that automatically helps me find cheap hardware on Government auction sites for my homelab. All from a single prompt lfg!
Computer use is finally reliable. It just needed a real harness. Now there is one.
Starting today, anyone can ship a SOTA browser agent, batteries included.
Introducing Browserbase Agents: one prompt and one API call is all you need to automate the whole web.
Being a great builder doesn't translate to being a great founder. You can much easily build great products working inside big companies where distribution and feedback loops already exist. You want to be a founder because you are slightly cuckoo in the brain.
Evaluating web agents on the actual web is hard. @VibrantLabsAI did it right: live Shopify stores, deterministic verifiers, all running on @browserbase. Open for everyone to run now ↓
1/n
For browser agents, a major bottleneck in evaluation is truthful scoring on the live web. A task is only as good as your ability to confirm the agent actually did it, on a real site whose state keeps moving and that the agent can potentially misreport.
So we took matters into our own hands.
Today, we're releasing Ecom Bench on @PrimeIntellect: 40 shopping tasks on real Shopify storefronts, each run in a live @browserbase browser and graded by a deterministic verifier.
https://t.co/NJ43WnTk4O
Platforms flexing a 90% "stealth benchmark" are treating users like clowns 🤡
Static snapshot tests aren't tech moats—web defenses change constantly under the hood. For real production scale, you need a partner built to survive a moving target, not landing page theater.
Your agents shouldn't relearn how to use sites every time they visit.
https://t.co/vjMb1zPyEL is the web skills catalog for your agents. Since launching 2 weeks ago the community contributed over 300 skills and 35,000+ CLI downloads.
We're 2nd on Product Hunt, help us win product of the day!
Building Browser Agents has never been easier.
Join us this Thursday (6/4) for an Opus 4.8 webinar with @AnthropicAI and @Letta_AI.
We'll discuss how we evaluate model capabilities with Stagehand and show a live demo on how to power your own agentic products with Browserbase.
Covers containerization, continuous integration, and automated deployment without overengineering it.
The best part? This exact setup powers the blog itself 🔄
Link: https://t.co/rwKfRBuNjT
More posts coming in 2025. Let's see where this goes.
#DevOps#Docker
🚀 Starting a new habit: writing technical blogs
First post on setting up a complete CI/CD pipeline—from "localhost:8080" to production with Docker, GitHub Actions, and automated deployments.
If you've ever wondered "how do I get this from my laptop to the real world?"
@alexalbert__@SullyOmarr Just a lot more simplification front and center most commonly used actions directly in your face. Reduce the rest directly cluttering the page, perhaps showcase the rest as icons or something else and explain more when hovering
🎊Big News! In collaboration with @googlecloud and @gcloudpartners, we’re launching DeepPCB Pro– the advanced version of our AI-powered platform for PCB design!🎯Join us on Sept 12th, 4:30 PM PDT at Google SF to be part of the future of hardware. RSVP here: https://t.co/h0lesJkGjo #AI #PCB #DeepPCB #PCBSolutions
The NT Family is growing! 🐣
✨ Introducing ChatNT, a Conversational Agent designed to analyse genomics sequences and address a wide range of key biological questions, assisting scientists in their daily work 👩🔬
📚Paper: https://t.co/gkHiEgN8rb
🌐Blog: https://t.co/ods98i8s9K
USA Immigrant Visas issued by Country (2023)
Mexico — 66.5k
Dominican R — 33k
Philippines — 27k
India — 25.5k (!)
Cuba — 23k
Vietnam — 22.5k
China — 22k !
El Salvador — 15k
Afghanistan — 12k
Jamaica — 10k
Bangladesh — 9.5k
Colombia — 9k
Nepal — 8k
Total (all) — 493k
1/3