Constantly frustrated and confused, purveyor of useless opinions. Fascinated by AI. Rancher, k3s has been. Member @Ch_JesusChrist

Phoenix, AZ
Joined July 2013
Darren Shepherd retweeted
We found a few tiny regressions the agent missed, mostly different shapes on buttons or slight ux regressions, overall I'm amazed how well this worked. Technical dept is can be counted in token now.
7
3
123
29,254
Darren Shepherd retweeted
Darren Shepherd (Obot's Chief Architect) wanted his coding agents to run the app, open a browser, test their work, and install whatever tools they needed. And he wanted to do all of that in parallel. So he built Discobox, a way to create and manage sandboxes that give coding agents their own machines. 👀 Learn more: youtube.com/watch?v=0ql6RGBu… @ibuildthecloud #Discobox #CodingAgents #AgenticAI #DeveloperTools
2
1
3
254
Darren Shepherd retweeted
this is how an elite anthropic technical member of staff prompts claude code btw
122
24
15
1,389
174,351
Darren Shepherd retweeted
1/ muse spark open is the most capable US open-weight model to date. 816B total params, 42B active. 1M-token context. natively multimodal. built for coding and agentic work like long-horizon tasks, hard software engineering, multi-step reasoning. america! 🇺🇲
298
461
222
5,602
681,235
Darren Shepherd retweeted
Replying to @alexandr_wang
coming for the open source market
1
54
7,175
Darren Shepherd retweeted
What if saying `claude build me a cloud platform make no mistakes` actually worked??? With the newest Antithesis release we have closed the loop and enabled agents to ship reliable systems: antithesis.com/blog/2026/tim…
5
23
12
156
30,587
Darren Shepherd retweeted
Source code is not available yet; it will be available in the coming weeks. Launching the website now to get the project qualified for @AnthropicAI [1] Glasswing overalls.dev/ [1] 🐜 folks, hit me up in DMs.
31
20
7
364
18,374
Darren Shepherd retweeted
Periodic reminder that you can buy yourself a CPU/RAM slice and run all the permanent VMs on it you need: exe.dev/pricing Make them public behind our GLB. If you want to know a little bit about the custom architecture, here are some new details: exe.dev/docs/architecture
9
5
148
15,628
Darren Shepherd retweeted
Discobox is coming up on The Context. 👀 Join our Chief Architect, Darren Shepherd (@ibuildthecloud), for another look at giving every coding agent its own machine. 🗓️ Thursday, Oct. 15 Set a reminder: buff.ly/xZqUqrT #MCP #AIAgents
1
2
3
1,210
Darren Shepherd retweeted
Rafael Ibayev
@rafaelibayevOct 11
Replying to @davidfowl
x is basically the nightly build. everyone else is on stable and quite happy there
1
4
2
63
4,003
Darren Shepherd retweeted
introducing msb in 90s, a new series. episode 1: pause and resume pause a running computer. resume it right where it left off.
3
4
4
61
14,283
Darren Shepherd retweeted
My co-founder, @ibuildthecloud, has been building a new project called Discobox the last few months, and I think it is pretty cool. It gives every coding agent its own machine: an isolated sandbox with its own desktop, browser and network. Agents can run wild in parallel without touching your laptop or your credentials. It's open source. We'd love any feedback if you give it a try. obot.ai/blog/meet-discobox-g…
1
2
6
378
Darren Shepherd retweeted
Darren Shepherd
@ibuildthecloudSep 30
Replying to @jamonholmgren
Skills are local specifications that help answer the "it depends" question. There's no universal right way, everything depends on context and skills and docs set context.
1
1
1
815
Darren Shepherd retweeted
"Why buy a sandbox? Just launch a container." @ibuildthecloud in the latest episode of The Weekly Dev's Brew Then the coding agent hit parallelism, colliding work trees, secrets, and egress. Sandboxing only the tool call is flawed. The agent belongs in the sandbox. 0:00 Why sandboxes felt dumb 1:38 Building a coding agent 2:36 Parallelism and work-tree chaos 3:25 What a sandbox actually is 4:27 Files, secrets, and egress 8:20 Tool-call-only sandboxing is flawed 12:28 Put the agent in the sandbox 13:31 Client-side vs centralized loops 14:51 GPT Script, Clio, and MCP 20:00 MCP, OAuth, and staying in lane 30:01 When AI skills get dangerous 35:00 The 100x developer myth 40:00 AI as a compiler 45:01 Distributed systems instincts 51:09 ADRs and keeping quality 55:04 Regression holes with agents 1:12:30 Cloud Code and tooling shifts 1:21:16 Wrap and finding joy with AI
8
2
2
15
4,313
Darren Shepherd retweeted
Breaking: Browser Use + Jev = Ultrafast ⚡ Findings flights took 7s and cost only $0.0039 🤯 > new action space every step > DOM state space > small LLM fallback to type (this video is at 1x speed btw) Built a tiny open source browser agent. try it below ↓
300
660
337
9,133
3,265,932
Replying to @ibuildthecloud
You can run it with local models already
love jev, but upset it - isn't open source? - it lack vision capability? We fixed all of that, introducing SimpleJev.ai A fully open source library which takes any HF model and Jev-ify it, with an API endpoint Now on github, and live in production at @FeatherlessAI
2
6
1,112