Joined January 2026
- Tweets 75
- Following 15
- Followers 698
- Likes 93
23 Photos and videos
OpenWork (YC P26) retweeted
ok sorry guys, but openai realtime is really dope
super scrappy demo of how it works inside of openwork
shipping tomorrow in the app
3
2
21
3,084
So will you take the $2M yes or no?
Are $2M worth of credits in OpenAI really worth it?
693
OSSCAR Scaling rundown #5 => #1
Continuing the OSSCAR Top 50 countdown for the Scaling category! š
#5 Noctalia ships noctalia-shell, a sleek, minimal Wayland desktop shell crafted for the Linux community that wants beauty without bloat
1
1
1
376
OpenWork (YC P26) retweeted
I should've stopped at that tweet, but I didn't.
There's this thing with agents where you're mostly on a roll gliding through stuff, but then you hit that one task. It doesn't get fixed instantly. "WHY you exclaim WHY" the dopamine doesn't hit quite the same when it's not working. So you keep digging yourself into a hole. You're 15 commits deep. You're not sure exactly what they all do , but hey that was "work" so you keep them. You circle through models, providers, until you realize that you've been prompting into oblivion for 6 hours striaght. Then you look outside, it seems the sky is blue today. "Good" you say. "Let me try one more prompt"
1
3
790
Can't see much, but looks good.
Day 3 of making OpenWork Labs Launch Week
Introducing: Handsfree ā control your computer with your voice.
It uses MCPs for deep app integrations and computer use as the escape hatch for everything else.
Still early, but feels like the beginning.
3
568
OpenWork please make this graph more sexy:
We just crossed 15k stars on github our daily signups 10x in May.
3
464
OpenWork (YC P26) retweeted
We just crossed 15k stars on github our daily signups 10x in May.
9
2
28
2,185
OpenWork (YC P26) retweeted
Day 2 of making OpenWork š
- New org onboarding screen ā sign in and immediately see what your team set up (providers, marketplaces, workers)
- Pick your default model right from onboarding
- Cloud vs local provider badges so you know where things come from
- Toast when new providers get added ā click to pick a new default
- Model picker redesign: provider accordions, recommended models, choose exactly which models show up in your session
- Can't accidentally disconnect cloud-managed providers anymore
- Clean sign-out that actually cleans up
2
2
18
900
OpenWork (YC P26) retweeted
The big one: shell customization.
You can now toggle sidebar, status bar, docs, feedback, model picker, browser panel, welcome page -- all from settings.
This is the foundation for white-labeling. Change the app name, hide what you don't want users touching. Locally today, cloud-managed soon.
1
1
2
156
OpenWork (YC P26) retweeted
AI providers/models got their own global settings section. Separated from General so it's not buried. Uses the same components as the Cloud dashboard.
1
1
108
OpenWork (YC P26) retweeted
Sign-in button is right in the status bar now. No more digging through Cloud settings to connect. One click.
Moved authorized folders into its own Permissions tab. Settings home is now a hub that links everywhere instead of a wall of toggles.
1
1
1
144
OpenWork (YC P26) retweeted
Workspace icons are dynamic now. Each workspace name generates a unique gradient using @paper shaders. They actually look good at 16px. Also replaced the loading shimmer with an animated orb -- way cleaner.
1
1
1
140
OpenWork (YC P26) retweeted
Been on a P2 feature bender. Shipped a bunch of stuff I've wanted for a long time in @getopenwork today:
1
1
10
511
Bunch of new features coming together:
- New OpenWork MCP
- Compatibility with Anthropic Plugins
- Marketplaces to assign specific plugins to your team
- RBAC
Just launched the easiest way to share your agents/skills/mcp with your team.
We built the OpenWork MCP to essentially allow you to write:
"Please share this skil/mcp with the sales team"
Sales team then gets access to it via a small marketplace.
It's our favorite way to share skills internally.
1
3
570
OpenWork (YC P26) retweeted
after playing around a lot with ralph wiggum loops i think i realized this kind of long running agent is only valuable for these use cases:
- research (experimental auto training)
- building new product from scratch
- processes that have time bottleneck (eg. long computation)
for other types of long running agents it's better to rely on scheduled agents (have 50 right now) that maintains a state across runs in a more non deterministic way
1
1
6
554
OpenWork (YC P26) retweeted
Apr 24
OpenWork (@getopenwork) is the open source alternative to Claude Cowork.
Roll out AI compliantly across your non-technical org at scale. Use your existing agents, deploy on-prem, and use any LLM provider you want.
They already have 14k stars on GitHub with over 150k downloads.
Congrats on the launch, @benjaminshafii!
ycombinator.com/launches/Q3pā¦
42
44
439
90,435
OpenWork (YC P26) retweeted
Very impressed by ChatGPT Image v2.
Fed a text-only transcript from an internal discussion about the @getopenwork roadmap.
3
5
20
2,805
OpenWork (YC P26) retweeted
Apr 21
Claude code can get expensive really quickly with api credits. Just spoke to an org spending $54k / month between 9 developers in the AI team.
This doesn't scale.
Btw, we already support any LLM at @getopenwork. You can get Sonnet performance at 1/10th the cost with kimi-k2.5.
3
1
16
2,211