OpenWork (YC P26) retweeted
Working on new benchmark for OPUS 5.5 vs Deepseek 4.1
Exploring various ways to safe costs, maintain quality, and, additionally, to do that with very little effort.
Techniques I'm exploring:
- Dynamic model selection based on skill availability
- Seamless ways to create skills to reduce costs
- Using Opus as main agents and deepseek subagents.
Early findings Opus + deepseek sub agents, is not efficient at all. More soon.
OpenWork (YC P26) retweeted
Asked Fable 5.2 to make a game about OpenWork in @getopenwork.
And yes this is pretty much what we help with
api.openworklabs.com/mcp/pub…
OpenWork (YC P26) retweeted
OpenWork Computer Use + Photoshop Clone = AI Graphic Designer
I asked OpenWork to create a vintage California postcard.
It explored San Francisco in Apple Maps, found the perfect view of Dolores Park, then opened Photoshop to turn it into a postcard.
All by controlling my computer
Full 8-minute walkthrough:
00:00 — Intro & the challenge
00:20 — The vintage postcard inspiration
01:00 — Exploring San Francisco in Apple Maps
02:20 — Finding the perfect Dolores Park view
03:00 — Planning the design with OpenWork
03:40 — Opening Photoshop & setting up the canvas
05:00 — Adding vintage California typography
06:00 — Refining the design & layout
07:20 — Final touches & the finished postcard
No scripts. No Photoshop API. Just OpenWork using a computer like a human.
OpenWork (YC P26) retweeted
Coffee over MCP
Try it at coffee.openworklabs.com/sf
OpenWork (YC P26) retweeted
Can DeepSeek V4.1 Flash match Opus 5.5 on real knowledge work?
I ran both on 50 of my anonymized Granola calls, turning them into one action log.
OpenWork (YC P26) retweeted
Prepping a video that compares SOTA closed source vs SOTA OSS on knowledge work.
I anonymized some of my real knowledge work and compared our real data to verify this against things that are actually useful to our team.
Early results: you can get to SOTA level performance with good generic skills.
OpenWork (YC P26) retweeted
Remote sessions are live in OpenWork.
Kick off a task on your laptop, send it to another signed-in OpenWork computer, and track it like a subagent.
OpenWork (YC P26) retweeted
Today we're releasing WorkBot and OpenWork in Slack.
WorkBot is a simple chat interface for people who don't know or care what a skill or an MCP server is. They just want to get their work done.
You set up skills, connectors, and models once, from any MCP client. They're then available everywhere your team already works: Slack, WorkBot, the OpenWork app, and even Cowork and Codex.
OpenWork is free for up to 5 users and works with any model provider.
OpenWork (YC P26) retweeted
Biggest misconceptions bout OpenWork is that we're a desktop app.
OpenWork is a way to deploy AI capabilities safely to your team.
Whether it's MCPs, AI Models, or skills - admins can configure them all in OpenWork and distribute them across a wide way of surfaces.
Different harnesses, slack, simple chat interfaces, or the desktop app are all supported.
OpenWork (YC P26) retweeted
Live Artifacts are live in OpenWork.
You can know create small dashboard directly from the chat. The OpenWork twist on "Artifacts" is that it can use any data from your MCPs and package them into shareable widget-like surfaces.
Check this example where I create a dashboard for a Notion CRM.
OpenWork (YC P26) retweeted
Hi All,
We're hosting a Demo Night this Thursday with
@primitive @getopenwork @freestyle_dev
Its our first event in months and we've got some cool people showing next gen agents + next gen boxes.
[🔗 below]
OpenWork (YC P26) retweeted
We invested a lot in custom testing tooling for OpenWork.
We can test complex setups across different OS and network conditions, and make it repeatable.
All in deterministic .ts code. Not agent-driven, but made for agents to write, run, and debug.
OpenWork (YC P26) retweeted
Enabling Astra to your team with Open work gives you something pretty special:
- You already have SKILLs and MCPs synced
- With your Desktop Policies you already have access nailed down.
- You have predefined Workflows that run incredibly fast across multiple MCP.
- You have visibility on Automations so cost is under control.
And I keep going...
I joined @getopenwork because I wanted to refined how work looks like. This is starting to look more like it.
OpenWork (YC P26) retweeted
You can now create Custom Dashboards right from within OW.
Custom dashboard takes any tools from the MCPs you connect to OpenWork and allows you to display rich interactive UIs. All built on standards (MCP Apps + Code Mode)
OpenWork (YC P26) retweeted
We shipped Workflows in @getopenwork
Workflows give you n8n-type determinism with the convenience of LLMs.
When you do some tool calls you can say something like "remember this as a workflow" and it will create a deterministic script that can be (re-)run.
We didn't re-invent the wheel this is possible via Code Mode pattern.
And it composes with everything else we've been building:
- Works from the Web or Desktop
- Works against automations (much cheaper to run automations)
Let us know if you find it cool
OpenWork (YC P26) retweeted
Introducing OpenWork Web
OpenWork Web gives you a computer in the cloud with the entire @getopenwork experience.
Just sign in and you get access to your entire OW ecosystem right there. All the skills, MCPs, and AI providers are instantly available.