Jezz

designer turned developer, no-code adopter, former Technical PM at HUB and Supra.

[Jezz @dejesigner]

AI is faster but we need the right process

[Jezz @dejesigner]

Before: cluttered hero, an AI-looking hand, "17+ years." After: a clean photo, the real workshop steps, "50+ years in stainless." Applied by my AI crew, reviewed by me, on staging for the client.

https://t.co/gmBAtCH0Ch

[Jezz @dejesigner]

let me know if you also love this feature of codex

[Jezz @dejesigner]

https://t.co/gmBAtCH0Ch

[Jezz @dejesigner]

Next idea for CrabFlow 🦀

Shelly, our AI designer, is moving from screenshots into Figma:

→ real frames from real components
→ variables synced with the code
→ edits the same frames on feedback
→ hands Pinch the frame links for exact specs

[Jezz @dejesigner]

Client feedback in Taglish. My whole instruction: "here's the doc, implement the feedback." Echo translated it and wrote a 6-request ticket with steps and acceptance criteria.

https://t.co/gmBAtCH0Ch

[Jezz @dejesigner]

AI agents that ask before they act. They run on your computer and your own Claude or Codex plan, and you approve from your phone, then ship. 🦀

https://t.co/GuujhVBKyw

#crabflow #aiagents #solodev #buildinpublic #devtools

[Jezz @dejesigner]

@kartikb753 Good catch, ticket text alone won't cut it. Echo already reads the code when it writes a ticket, so the idea is to have it map likely files and follow imports up front. Then still watch the real diffs as crabs work, and stop or combine when they overlap.

[Jezz @dejesigner]

My AI agents keep editing the same file. So I'm hiring them a boss: an orchestrator agent that watches every ticket and says "stop" or "combine into one" before the conflict happens. 41s: https://t.co/KmSGppPyaN

#AIAgents #BuildInPublic

[Jezz @dejesigner]

@DmitryBBLV Agreed. The conflicts I hit weren't really git problems, they were timing and ownership problems. An orchestrator that says "you take this file, you wait" is the next piece I'm building toward.

[Jezz @dejesigner]

@dogramaci Honestly: give it a messy, real bug and see where it struggles. Agents still need you to review the PR and approve commands, and parallel crabs can conflict on the same files.

Here the invite link to send bugs https://t.co/eyBzOQTH1X

[Jezz @dejesigner]

Building alone with AI agents? 12 terminals, scattered bugs, zero focus. This is why I built CrabFlow. 🦀

https://t.co/GuujhVBKyw

#crabflow #aiagents #solodev #devtok #buildinpublic

[Jezz @dejesigner]

@corhoonx Haha it really was. I spent more time arranging terminals than shipping 😅 Now each crab gets its own branch and I just watch one board. Thank you! 🦀

[Jezz @dejesigner]

18 years after graduating, a recruiter asked for my Transcript of Records for the first time.

Pulled it out. Artificial Intelligence: 1.6.

2008 me had no idea. 😄

[Jezz @dejesigner]

Client: "urgent revision." Me: laptop at home, phone only. Pasted their doc into a CrabFlow ticket and let the agents plan it. Wrong grammar? AI gets it anyway.

https://t.co/gmBAtCH0Ch

[Jezz @dejesigner]

Type a note. Echo writes the ticket, Atlas plans it, and 8 crab agents build it, each on its own branch. 🦀

https://t.co/GuujhVBKyw

#crabflow #aiagents #claudecode #devtok #productivity

[Jezz @dejesigner]

Client: 2 photos + "please add this new vehicle". Me: forward it from my phone. Echo (my Prompt Engineer agent): full ticket with goal, context, files, steps, out of scope and acceptance criteria. 40s: https://t.co/KmSGppQ60l

#AIAgents #BuildInPublic

[Jezz @dejesigner]

https://t.co/GuujhVCio4

Built for solo devs, freelancers and small startups. #buildinpublic

[Jezz @dejesigner]

Too many terminals, an AI agent in each one, and you in the middle.

CrabFlow gives your agents one board. Write a note, and a crab crew plans it, builds it on its own branches and hands it back for review. It runs on your own Claude or Codex plan.

👇

[Jezz @dejesigner]

16 AI agents, 16 tickets, 1 starting commit. 16 open PRs = 120 ways to collide, and 5 PRs touched the same file. 45s on where my agents went wrong: https://t.co/KmSGppQ60l

#AIAgents #BuildInPublic