Founder of Skyvern (https://xitter.gq/t.co/aRwSa7swlf)
Joined July 2024
- Tweets2.8K
- Following112
- Followers484
- Likes626
Pinned Tweet
1/ We just raised $2.7M to give AI "hands" to do real work. 🚀
Browser automation has been broken for 20 years. If a button moves 10 pixels, your script breaks.
Meet Skyvern: The open-source agent that fixes the "maintenance tax" of automation.
Why is this so expensive!!!!!
Today we’re launching the Gemini Decisions API, powered by our new Gemini Decision 1 model in @GoogleAIStudio.
Instead of generating text token by token, it’s built for closed-set choices scoring up to 1,000 questions across text, images, audio, video + PDFs in a single pass. A few details:
- 3 decision types: yes/no, multiple choice (up to 255 options) + rubric scoring
- $0.10 / 1M input tokens with $0 billed for output tokens (~5x cheaper than Flash-Lite)
- leads on several benchmark evals with lowest calibration error
Google teams are already building with it, below is a look at Stitch Rapid designing a full product flow in seconds (1x speed)
32
Things I worried about when I started an AI company:
- Will AI get too smart
- Will it take everyone's job
- Is it safe
Things I worry about now:
- Am I absolutely right
- Is it lying to me
Same here. This is taxation, except some of the taxes go to shareholders for unknown reasons
Our health insurance is going up 25% for the exact same coverage next year with zero changes.
Happening all over the US. Where health insurance is covered employers are just swallowing it; employees don't even see it.
66
Cant wait to try this out in @skyvernai
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
52
FWIW I did this with our BI tool and have never been happier
I am no longer bound by their opinionated design
My mind is being blown by some of the responses to this post?
Do people really think building your own in-house version of say something like Airtable is worth the $600/year that you end up saving?
Is there nothing better you could be doing with your time? No disrespect but wtf lol?
40
You don't have these problems w @skyvernai
PSA for those of u vibecoding naked on your laptop with no containers: tell your agents to DELETE Playwright NOW. Even the latest (1.64.1) bundles early Chromium.156 with
CVE-2026-118306
CVE-2026-118331
CVE-2026-118348
CVE-2026-118344
And the Playwright SANDBOX IS OFF BY DEFAULT
1
98
Every engineer that hates it isn't thinking beyond their slice of the world
When I first told @MParakhin that we were planning to move to a monorepo he said wisely:
Moving to a Monorepo is an idea that every engineer absolutely hates, and that once its there, no engineer would ever want to work at a company that doesn't use one. We will definitely do it.
52
The internet was FUD before it wasn't... last year may have been a canary in the coalmine
The AI jobs panic peaked in May and has faded fast since:
- News reports on AI job loss are down ~65%, back to year-ago levels
- Layoff announcements that cite AI are down ~90%, near the lowest of the year
Charts of the Week: a16z.news/p/charts-of-the-we…
70
🚨 🚨 WEBINAR 🚨 🚨
We're doing Skyvern's first WEBINAR this week at 10am ET! COME JOIN!
Link to sign up below
Progressive disclosure is 10x better than RAG
We rebuilt Skyvern to use it and it got 2.3x faster and 50% cheaper
skyvern.com/blog/deleting-ra…
Anthropic is now recommending that companies REMOVE unnecessary subagents from their AI systems.
This is a pretty significant shift in agent architecture, with their team finding that a single agent equipped with skills can frequently outperform complex multi-agent systems, and one customer seeing roughly 2x faster execution.
The emerging rules of thumb:
→ Skills: Teach the agent HOW to do something, including procedures, expertise, and instructions.
→ Tools / MCP: Give the agent the ability to DO things through APIs, actions, and live data.
→ Subagents: Use for parallel work, isolated context, or tasks a single agent struggles with.
The interesting part is progressive disclosure. Instead of spawning specialized agents or stuffing everything into a giant prompt, one agent discovers and loads skills and tools as needed.
Their advice is to fold existing subagents into the main loop, turn their instructions into skills, expose their actions as tools, and run evals to see if performance holds up.
As models have gotten smarter, improving agent architecture is becoming less about what you can add and more about what you can delete.
claude.com/resources/article…
3
108