ChicagoDave

The number of times those get grabbed of someone’s face and stomped on will be greater than zero. And businesses will have signs for No guns/spyglasses.

I’m not sure the message should be benchmarking.

The eye-opener is clean licensed data with filters for AI content (not sure how you do that).

If MSFT builds up using an ethical approach, there is a large anti-AI audience that might take note.

Yeah this is exactly what I was thinking.

Vibe coding is the chaos version of using GenAI. Seven agents in parallel and the dev refuses to believe this will just be a circle jerk of hallucinations.

Planned, architected problems led by experienced dev/arch’s will be where success is measured.

I've been thinking about this since ChatGPT arrived.

My thoughts are distilled in a single page:

https://devarch.ai/trio-paradigm.html

I doubt we will ever need a large team to build software again. PMs will be fractional. Product owners and subject matter experts will become more valuable. And engineers that have deep experience with things like Domain-Driven Design will thrive and vibe coders will eventually be shut out of bigger roles.

I'm the OP.

I started using ChatGPT and Claude, then Claude Code the second each tech stack became available. I iteratively cajoled as much productivity as I could from it. Until October 2025 I was adamantly against it as a real tool for complex software projects.

Then Opus 4 -> 4.5 arrived and as we all know, changed the world.

But during those earlier times with minuscule context windows and many hallucinations, I still extracted evidence of what could be accomplished. I stored over a thousand work summaries from several projects and then in January 2026 I worked with Claude Code to build DevArch on the principles it could discern from those work summaries. Obviously I corralled Claude into developing a set of guardrails that I felt encapsulated what I describe as productive architecture and software engineering from a continuous improvement stance.

DevArch is now at 4.0 and installs on any Mac, Linux, or WSL OS. There's a 14-day free trial so you can download it and see for yourself.

If you're serious about being productive and generating high quality code, then I believe DevArch is a great asset to your toolkit.

If you're a strong Claude Code/GenAI dev, then you're probably not my target market. If you're a dev that finds themselves constantly re-prompting the same things over and over, discovering empty unit tests, fake methods, hours of wrong direction work....then you are my target audience.

And "my workflow" isn't a workflow. It really allows you to work however you want...with DevArch watching everything and injecting questions when it senses you're straying from your mission. And it will push you to wrap up, commit and push, clear the deck, then recap and repeat.

Nice idea, but engineers are engineers for a reason.

I’d suggest that the domain expert partner with a GenAI senior engineer to build together. In fact I believe this is the new dev team model. Domain Expert + Senior Engineer + QA. Not sure we still need a project manager anymore and we certainly don’t need scrum masters.

I do mostly cloud modernization architecture work and use Claude Code daily. But even I’m burning out on it. Not sure if that’s my current job, GenAI, or that I’d rather be playing poker.

I have five Gen-Z kids that are pushing back on GenAI hard and they claim their peers are getting angry about how it’s impacted their lives.

As a technologist I tend to lean into new things rapidly because that’s how I’ve survived in IT for so long. Since I’m not ready to retire I still have a vested interest in staying informed.

But the OP has definitely identified a psychological issue I think we’re all going through.

I’ve started pumping the brakes on Claude usage. Before I would invent a target to work on. Now I’m filtering existing tasks to needs and not spending nearly as much time in Claude.

I’d bet this is being felt by the AI companies and the correction we’ve been talking about is nearing.

GenAI is great as a tool. But it can’t be everything.

This is how I do it to and I’m an architect/senior dev. Keep it up!

I just read through your comment history.

You literally hate everyone and everything.

I couldn’t find a single instance of positivity.

It almost seems like you hang out on HN just to scold people.

You never actually engage in conversation.

You probably have a hit list and automatically down vote everything they say.

100% agree with this article. You will always need to lead the vision and architecture.

Maybe the link was misplaced but the original comment stands.

I played Dungeon, the much larger mainframe version of Zork in 1978. Between me and six other guys it took us nearly a year to finish the game.

At that time, there were no video terminals. There were no monitors (this is high school).

We played Dungeon and Adventure on 17” wide green bar paper terminals, usually a Digital Equipment Corporation DECWriter II or III.

There was no Internet. There was nowhere to go for hints. We simply had to figure everything out.

At that time, these were the first complex computer games. When the Imps created Infocom they made the top ten most popular games until video arrived in classrooms and homes.

There is a fairly active community of hobbyists that still make text games, though evolved away from the brutal puzzles to more balanced narrative and seamless puzzles.

Using Claude I even built my own platform: https://sharpee.net/.

There a thousands of free text games. Check out https://ifdb.org for more.

There are individual tasks that can be “swarmed”.

Constructing software requires supervision and a continuous feedback loop from a discerning human.

I’m just a guy with 41 years building Fortune 500 software. What do I know, right?

When you unleash it without a plan or discipline, they are correct.

With discipline, it’s an aggregator.

https://devarch.ai/

Even with the strongest guardrails, no llm can be left unsupervised. And believing an llm can supervise itself is madness. For every agent you deploy as a “team” you will have multiplied diminishing returns.

I’ve iterated the process hundreds of times and even with a strong specification and tests, an llm can meet the spec and pass all the tests and still build the wrong thing.

This multi-agent workflow idea is worse than web3.

But it’s a YC thing so someone will make money.

Still misses entirely the way software is constructed. These whack-a-mole tools are pointless.

Finally, a post that explains the difference between skill+GenAI > anyone+GenAI.

One is augmentation and the other is whack-a-mole.

Excellent read.

Pretty sure Americans hate him too. You can see Tesla new car staging lots full and the cars never move. They try to hide the Cybertruck lots.

Sigh.

I’d hoped Anthropic would steer clear of blatantly unethical practices but here they go in bed with that guy and his horribly damaging data center.

Another post that doesn’t understand effective use of GenAI in software engineering.

The assumption is that there’s no way to extract speed and accuracy matching business models.

This isn’t obviously false to the majority of dev/arch’s because most are vibe-coding, but it is extremely obvious to the minority that has focused on accuracy first THEN speed.

https://devarch.ai/

If you're using GenAI, you should go through the process of selecting an optimal tech stack for each solution, but also take into consideration that Claude and other services probably the most knowledge of python, javascript, and typescript with go, rust, java, and c# following closely behind. Consider what you're building and what elements of the tech stack is optimal for your problem-space.

I don't know rust at all and I've built three applications using it with Claude because it has speed and correctness built-in.

I use Typescript for 90% of the things I build. For web development I've used a number of tools, but mostly react, nextjs, or raw html/css/js. But if I were building an enterprise application I'd consider my team and whether opinionated (Angular) was optimal over flexible (React).

Each project should consider its own optimal tech stack.

I encourage skeptics to look at the Sharpee repo at https://github.com/chicagodave/sharpee and specifically docs/context. Every session summary is there going back months.

I have not discovered anything new. I have applied established architecture and engineering practices that I've been following for 15 years to using Claude Code.

Domain-Driven Design started with Eric Evan's "blue book" in 2003 and continued with Vaughn Vernon's "red book" a few years later. Event Storming workshops and event sourcing data storage have also emerged as important tools.

I am just a practitioner. Eric, Vaughn, Paul Reyner, Alberto Brandolini, and a few other architects lead.

My DevArch guardrails is a toolkit to follow these practices.

I’m doing it because as a DDD adherent, I’ve been building software that way for 15 years without GenAI and now with GenAI I can do it faster.

You can’t play whack-a-mole with GenAI. You have to start from well-known principles and watch everything it produces. Every module or bounded context has to have its own invariants.

You can’t fully automate software engineering with GenAI. It seems the vast majority of GenAI users think they can and end up in the same place as the OP.

Maybe learn Domain-Driven Design, Event Sourcing, and then try again. The results will be dramatically improved.

https://devarch.ai/

Am I the only one that read Greg Bear’s novel Blood Music?

That book has haunted me for decades.

I’m convinced my love of grocery store checkout logic puzzle books are the foundation of my programming skills. I was also good in tenth grade geometry which is all logic.