My AI workflow evolved from prompts to a near-autonomous workflow
12 comments
I wrote the full breakdown here: https://codeaholicguy.com/2026/04/11/how-my-ai-workflow-evol...
Sounds great, what I wonder is how you treat edge cases that your Ai requirement phase identifies. Do you let Ai decide how to handle them or do you jump in and review/adjust manually. I'm currently getting my hands dirty with spec kit and feel like I write very detailed specifications. Still edge cases occur regularly where I then adjust the spec or plan.
In my requirement review and design review phase, I specific that the agent need to review edge cases and come up with 3 suggestions. So that I can choose. In case it is not match my expectation I will tell it to remember the case and provide answer for it to remember.
Nice, maybe I'll give devkit a try to compare it to the spec kit approach.
Appreciate it, I know there are so much room to improve, looking forward to hearing feedback from you.
i have a similar workflows for bugfixes with claude but threaded together manually.
1. identify bugs 2. ask claude code to investigate the codebase for three potential root causes. 3. paste the root causes and proposed fixes to claude project where i store all architecture doc and design decision for it to evaluate 4. discuss with claude in project to write detailed task spec - the task spec will have a specified format with all sorts of test 5. give it back to claude code to implement the fix
it works most of the times. and i think the key to a successful workflows is that it allows human involvement at the critical moment like product decision, verification of proposed fix, so that model wont just fking freestyle and hallucinate.
1. identify bugs 2. ask claude code to investigate the codebase for three potential root causes. 3. paste the root causes and proposed fixes to claude project where i store all architecture doc and design decision for it to evaluate 4. discuss with claude in project to write detailed task spec - the task spec will have a specified format with all sorts of test 5. give it back to claude code to implement the fix
it works most of the times. and i think the key to a successful workflows is that it allows human involvement at the critical moment like product decision, verification of proposed fix, so that model wont just fking freestyle and hallucinate.
I just want to say i appreciate you wrote the article and do your own writing. I dont really understand what your trying to say (something analogous to omc/ulw?) but i appreciate it
Thank you, appreciate it
Great question! I have been thinking about this too and would love to hear what approaches others have tried.
The shift that mattered most was not better code generation. It was moving from reusable commands and templates into a workflow that can carry context, trigger the right behavior, and verify work automatically.
A recent feature made the difference obvious. I used Codex to build interactive skill selection for "ai-devkit skill add". I gave one sentence of instruction, and the workflow carried the task through requirements, design, planning, implementation, verification, tests, and code review.
The whole session took under an hour. The actual feature flow was around 30 minutes.
What I found interesting was not just that AI wrote code. It was that the workflow left behind requirements, design docs, planning artifacts, tests derived from requirements, and verification against the spec instead of just a diff.
A few things that felt important in practice:
- memory pulled back an old CLI rule I had forgotten I stored - review phases could loop backward instead of blindly moving forward - verification caught drift between implementation and design - I still made the product decisions and fixed the last failing test myself
I am curious how others here are thinking about this.
Are you mostly optimizing prompts, or are you now trying to optimize the workflow layer around the model?