thisdougb

What has always worked for me is taking small steps and being outside. Even just go outside once, to the shop further away. Then do then do that again. Small and near-term goals. Today, this afternoon, in the next 10 minutes.

What has always kept me in that funk is trying to get back to 'normal' as a goal. It feels too big a jump, so I'd put it off for another day. And then a week has passed.

I don't want this to come across as a plug. But I made an app for myself in these situations (and to avoid them). Check my profile and you'll find it. Message me somehow and I'll give you a free subscription, in return for some honest feedback.

But small steps, and be outside more. As others are also saying. Just witnessing life is carrying on as normal really is refreshing.

Very impressive. But:

"When agents write code, the number of changes goes up, fast. In Anthropic's monorepo, changes landing per day went from about 2,500 in May to more than 30,000 in September this year."

changes == bugs ?

Aren't they (CTO, Engineering Lead) questioning why that volume (increase) of changes is required? Is Claude Code (or its use) just eating itself?

I'm always puzzled by the claim 90% of code at Anthropic is done by Claude Code. My own attempts to coax it into coding anything stable just ends up with loads of small bugs (I don't count docs as code). They're setting this expectation, amplified by AI evangelists, but the reality is far away for most.

Does any have examples of AI created apps or web apps that are a success? Genuinely want to use/follow them to see if we're past the hype stage, and it's just me that's getting it wrong.

I just use a branch per feature or unit of work. The git hooks make the time measurement automatic.

  git checkout -b feature_a
  <do work>
  git checkout master
  git merge feature_a

probably about 10 seconds of effort, over the course of 1-3 days of coding time.

After quite a few years of coming up with and implementing 'great ideas' but not being able to follow through to making them revenue generating products, I'm on my best bet so far.

I always wanted to build a real-life puzzle game, which is app/mobile assisted. Had yet another eureka moment, and built a usable prototype (backend plus iOS app). Good feedback from a small circle.

For a while I was aware of someone (I knew by sight) who worked in the same sort of subject matter (but a non-tech). I approached her, we had a coffee, I pitched the idea and how she could bring it to life, as I made the tech side. She jumped on board.

We're two and a half weeks in, have gone full speed and are making something great (for our audience). My future co-founder is amazing, great insights, opinions, drive. We're potentially launching in a couple of weeks, a free/MVP version of a puzzle game.

I've been through many iterations of trying to get something off the ground. Tried tech co-founders, and the last years of going solo (very hard after you've done the coding). But this now feels right. A puzzle app/game for every day people to have some fun. And a future co-founder whose life is outside tech, who's bring a sort of fun energy outwith let's make loads of money or isn't the framework/AI cool.

Balance is good. Contact with reality is good too :)

I do a lot of code review and doc/comment updates on my iPhone, using Working Copy and a small bluetooth keyboard. It turns out to be ideal because the small screen size means better mental focus.

I have a vm I connect into via ShellFish (same dev as Working Copy) for when I want to checkout/fix/deploy, but it's a rarity that I need to do that.

Why does it matter if they're called posts, pages, or bananas?

Static site generators take input, run it through templates, and produce html output. This seems to be, literally, what you want to do. Maybe try and implement what you want and you'll see they can do it.

I use Hugo to generate some static pages that sit inside a dynamic website. It's just Go templating, which is not too hard to learn. You can create individual pages that pull in multiple templates if you want. You can switch off RSS, change naming from posts to bananas, etc. Very few preconceived regulations.

Most of the static generators are used for blogs, that's probably why most examples they give are around blog posts. It doesn't mean the name 'post' actually means anything special.

That's a good read, thanks.

I'm really interested in the additional time cost of your switch. As you say, measuring productivity is seems nearly impossible.

But I would expect your 40 developers now have more time to do other things, because AI is taking on some of the workload? So have you noticed developers now doing 'new' things, like experimenting with new ideas, shorter hours, more holidays, or spending more time learning non-core topics? I think that would be visible. The 'Friday for personal projects' becomes 'Thu and Fri for personal projects', for example.

When frameworks became popular ten or so years ago, developers made the 'this will be much more efficient' argument. All that seemed to happen was the time they used to spend thinking/learning the core language (or engineering concepts) was swapped for time learning the framework. So there was no significant productivity gain, despite tasks feeling easier to do.

You list AI management/maintenance tasks that didn't need to be done before. I assume not all of these are 40x across your developers. It will be interesting to see if this management workload becomes similar to infrastructure management, and develops into a new job type providing a shared service. At least maybe during the transition away from flesh-based development to AI development.

I guess your end goal is that AI writes all (literal) code and owns the codebase, directed by people in the company? Do you have a plan to transition developers away from tech roles, is that something that's discussed?

This reads like mentoring a junior dev. You do mentoring because the pay off is a junior that develops into a senior; writes better code more efficiently. But what's the pay off with going through this process with AI?

A few days ago I spent an hour trying to get Claude to write unit tests for a relatively simple function, without getting anything functional in return. It produced good (but changing) plans of work, but the code generated didn't match the plan (syntax or semantics).

Are people simply replacing 'time spent coding' with 'time spent mentoring AI'? And is that a useful trade off, given AI forgets everything?

I'd love to see a more honest discussion about the overhead of coding with AI assistance. It's still too high an overhead for me at the present capability.