straydusk

Pretty unpopular influencer if that were the case

That's me! I'm pretty open about that.

I'm not trying to express that my particular flavor of career is safe. I think that the ability to produce software is much less about the ability to hand-write code, and that's going to continue as the models and ecosystem improve, and I'm fascinated by where that goes.

Ouch lol.

Is the image really not that clear? There are IDE-like tools that all are focusing on different parts of the Spec --> Agent --> Code continuum. I think it illustrates that all right.

I completely agree in a sense - the cost of producing software is plummeting, and it's leading to me being able to develop things that I would never have invested months in before.

I mean, fair enough. Obviously there are different levels of criticality in any production environment. I'm building consumer products and internal tools, not safety-critical systems.

Even in those environments, I'd argue that AI coding can offer a lot in terms of verification & automated testing. However, I'd probably agree, in high-stakes safety environments, it's more of a 'yes and' than an either/or.

I think people attacking "don't read the code" are not considering the status quo - they're comparing to some perfect world where staff engineers are reading every line of code. That's not even close to happening. Test-driven development is something most engineers just won't put up with... AI's will do it, no problem. If I can automate ten different checks for every commit, is my code really getting looked at less?

"Testing ladders" is a great framing.

My approach is similar. I invest in the harness layer (tests, hooks, linting, pre-commit checks). The code review happens, it's just happening through tooling rather than my eyeballs.

Exactly this. The code is an intermediate artifact - what I actually care about is: does the product work, does it meet the spec, do the tests pass?

I've found that focusing my attention upstream (specs, constraints, test harness) yields better outcomes than poring over implementation details line by line. The code is still there if I need it. I just rarely need it.

No pun intended but - it's been more "vibes" than science that I've done this. It's more effective. When I focus my attention on the harness layer (tests, hooks, checks, etc), and the inputs, my overall velocity improves relative to reading & debugging the code directly.

To be fair - it is not accurate to say I absolutely never read the code. It's just rare, and it's much more the exception than the rule.

My workflow just focuses much more on the final product, and the initial input layer, not the code - it's becoming less consequential.

It was extremely easy to see them coming, because he talked about the repeatedly.

The markets priced in him backing down repeatedly, which he has.

Location: Seattle

Remote: Prefer hybrid, remote for the right oppportunity

Willing to relocate: No

Technologies: SQL, A/B Testing, TypeScript, JS, Python, SQL, R, Next.js, React, A/B Testing, CI/CD, LLM Integration (OpenAI, Anthropic Claude), Vercel AI SDK, Prompt Engineering, AI-Assisted Development (Claude Code, Cursor), ML Enablement, Platform & Infrastructure (event pipelines, data lakes, warehousing), Governance & Metadata Catalog, Privacy Management (GDPR, CCPA), dbt, Tableau, Amplitude, Snowflake, Airflow, BigQuery, Athena, S3, Iceberg, Redshift, Dagster, Fivetran, Sigma, Mixpanel, Optimizely, Kafka, APIs

Résumé/CV: https://docs.google.com/document/d/1bK6v3M2YyNEHaOfREeso998m...

Email: ben.shoemaker.xyz@gmail.com

I've been in product management / leadership for the last 10 years - I've spent the last 6 months exploring the AI coding space & AI in general. I'm currently the fractional CTO for a startup in the real estate space; I've got a number of my own projects I'm working on; and I'm very interested in and involved in the developer tools space - here's an example (https://vibescaffold.dev/) from a couple of months ago that was featured in the AI Tinkerers Community Spotlight.

Yo, mind explaining your setup in a bit more detail? I agree completely - I like the Claude Code harness, but think Codex (the model) is significantly better as a coding model.

I'm struggling with landing in a good way to use them together. If you have a way you like, I'd love to hear it.

I finally found the perfect way to describe what I feel when I read stuff like this.

Assuming that's a response to me... that explanation would have been nice, and it's misleading without it.

What is this? How does it work? How are skills ranked? Seems a little bit fishy to me that you can only tell it's from Vercel if you click the top left corner, and the top two skills come from vercel... despite there definitely being much more used skills in the overall AI coding ecosystem.

The UI looks nice, otherwise. I had thought about building something like this - maybe this just increases my confidence that this is needed, just not affiliated with a company.

There are HOSTS of dogshit devs that operated that way, trust me. Half the job of a PM has been to work with these types of people.

This is so well-done - kinda surprised it's not getting more traction!

https://hn-wrapped.kadoa.com/dang

Sorry but you linked a source that is bogus be definition

Nice! Gonna give it a try for my new project.

Some thoughts: 1) I think you could simplify the onboarding - I don't think I actually need to make an account up front. It somewhat contradicts the way the landing page presents things. 2) The AI prompt does a good job - I really like the output it provides. 3) It doesn't seem like there's a super straightforward way to separate local & production logs - would be nice.

I implemented it - I'll see how it cooks the next couple of days & check back.

Side note, if you want to try that project, since you're a vibe coder, I'd appreciate the feedback too - vibescaffold.dev.

Added a new modal for this - I really like it! Thanks for pointing that out - it was a bad experience before! https://imgur.com/a/VLbjuW5

I wish it did as well. I am sure someday, this tool won' t be needed. Until then... maybe it will!

Thanks a lot! I agree (obviously) - this workflow really helped me structure my projects.

> If you're looking for feedback, you could include a tiny section on the homepage about how to run the output docs. e.g. put them in a folder, point Claude Code/Codex to it and give it the prompt.

This is great feedback - I totally agree - going to add this asap

Hey there, I changed the messy framing. Hope you like it!

I also fixed the wizard text, I agree.

I also fixed the step jumping.

Would love to hear your feedback if you try it again with the fixes in place :)

I fixed it, and a few other issues. It was because I hit my limits in the API... OpenAI billing is weird.

Thanks for saying this.

Uh... I have no idea why the chat isn't working! Looking now!

I think the reverse Socratic method when it comes to LLMs is one of the best constructs, period - both in this context, but in many.

Shot you an email!

I think you're right. I think that's why I like this approach - it's keeping you in the loop for most of it - and you just get documents as output that you can edit, check, and iterate on yourself.

I found SpecKit to be over-engineered, and BMAD to be quite complex. To me, this was a "thin" enough solution that I still felt like I was in control, and not at the mercy of an additional abstraction.