Nostr notes by Jason Gorman

We're very much at the "Ah, but what if it didn't ...

2026-10-30T09:07:56Z

We're very much at the "Ah, but what if it *didn't* have to be reliable?" stage of this AI journey.

What makes software engineering so challenging - more often than ...

2026-10-24T08:31:09Z

What makes software engineering so challenging - more often than not - is people.

I've worked with a bunch of technically very capable devs who routinely failed because they couldn't do the people part.

You can't learn to do it on your own.

LOL at employers. You'll very probably be needing those ...

2026-10-22T09:27:42Z

LOL at employers.

You'll very probably be needing those experienced engineers in the future. Probably even more of them.

And you'll very probably regret not hiring and training entry-level engineers, and that nobody else did, either.

But, as always, you won't quite join those dots.

I've been asked by several clients over the past year to help ...

2026-10-14T07:12:15Z

I've been asked by several clients over the past year to help them find senior software engineering talent. You need to play down the "AI" rhetoric if you want to attract people who genuinely care about the work and who really know what they're doing.

Advertising that you intend to install them into some kind of "reverse centaur" role will pretty much guarantee they won't be applying.

"Humans write bad code, too!" exclaim AI boosters who ...

2026-10-09T13:32:42Z

"Humans write bad code, too!" exclaim AI boosters who don't perhaps understand that's not an argument in favour of AI-generated code.

Sure. If you're gonna push bad code, at least make sure nobody read it first.

OpenAI accuses a rival Chinese lab of "stealing" ...

2026-10-05T15:07:18Z

OpenAI accuses a rival Chinese lab of "stealing" "hidden reasoning" by launching thousands of queries against their models.

That's quite revealing about what LLM "reasoning" actually is.

Pooled together the best available research and added it to a new ...

2026-10-03T13:29:48Z

Pooled together the best available research and added it to a new section "Code Maintainability & LLM Performance"

The 101? Don't listen to the folks telling you it doesn't matter if AI-generated code is maintainable by humans. The data says it really does matter.

https://codemanship.wordpress.com/2026/08/12/ai-software-development-what-does-the-data-say/

"Waterfall is the appropriate tool for any environment where ...

2026-10-01T14:44:52Z

"Waterfall is the appropriate tool for any environment where benefits realization is clearly understood, the final design can be engineered up front, and job steps can be sequenced before work begins."

Otherwise known as "never".

I think there's a good case to be made that folks using LLMs ...

2026-09-30T05:29:16Z

I think there's a good case to be made that folks using LLMs to do things they already have tools for that are quicker, more reliable and require many orders of magnitude less compute have lost the plot.

See also: people explaining to an LLM what code they want when it would actually be quicker to just write it.

Enjoying listening to my own stuff. That's the first sign of ...

2026-09-26T19:30:38Z

Enjoying listening to my own stuff.

That's the first sign of madness.

https://apeswithhobbies.bandcamp.com/album/love-songs-from-a-gilded-age

The problem with trying to get an LLM to follow a process is that ...

2026-09-24T07:31:15Z

The problem with trying to get an LLM to follow a process is that LLMs don't follow processes.

Thanks for coming to my TED talk.

"Gone are the days when developers just wrote syntax" ...

2026-09-11T05:21:54Z

"Gone are the days when developers just wrote syntax"

Yeah, I miss not having to think about semantics.

If you're interested in the design of agentic workflows, ...

2026-08-26T08:27:03Z

If you're interested in the design of agentic workflows, remember this formula:

R = 1 - (1 - P)(1 - C)

"Have you seen Anthropic's AI-Native SDLC playbook?" ...

2026-08-23T10:33:27Z

"Have you seen Anthropic's AI-Native SDLC playbook?"

No, but I've seen the Claude status page and the Claude Code issues page and I've read reviews of the leaked source code.

I think I'll be looking for advice on software engineering elsewhere, thanks.

I've worked with and spoken to hundreds of developers over ...

2026-08-18T06:47:26Z

I've worked with and spoken to hundreds of developers over the last year, and my best guess is that the true extent of AI use in software development is being massively exaggerated.

I'm determined to get to the bottom of this before the year's out.

What if I told you there's a way to 10x the productivity of a ...

2026-08-18T05:51:05Z

What if I told you there's a way to 10x the productivity of a software development team without making any improvements to their process?

Tackle problems that are 10x as valuable.

Strangely, that option is rarely on the table.

My London-based client is looking for a senior software engineer ...

2026-08-17T19:43:32Z

My London-based client is looking for a senior software engineer with XP experience on the JVM - Kotlin preferred but not essential.

£90K-£120K. Hybrid 2 days a week on site.

Not a bank. Not a blockchain something something. Established start-up with actual customers like in the old days.

OSS tools that developers use.

An established start-up based in London that makes super-useful ...

2026-08-13T12:45:07Z

An established start-up based in London that makes super-useful tools for software developers (so not a bank for once!) - is looking for a senior software engineer.

XP skills essential, and Kotlin experience is ideal, but experience doing JVM-based development is acceptable.

They'll consider less experience for exceptional candidates. I know the folks involved, and you'll likely learn a lot.

Salary: £90K-£120K, hybrid: cpl of days/week near Liverpool Street.

Ping me if this might be you.

When you take an evidence-based approach to AI-assisted and ...

2026-08-12T17:51:51Z

When you take an evidence-based approach to AI-assisted and agentic software development, a picture emerges.

It doesn't serve the narrative the bubble's been built on.

But for the tiny % who care what's probably real and what actually works...

https://codemanship.wordpress.com/2026/08/12/ai-software-development-what-does-the-data-say/

Bear in mind when you ask an LLM to explain why it did what it ...

2026-08-07T06:47:31Z

Bear in mind when you ask an LLM to explain why it did what it did (or didn't), you're asking faces on toast to explain faces on toast.

It's not explaining. It's plausibly completing.

And you fell for it twice.

Germans have a big advantage in those "Can you describe in ...

2026-08-04T21:11:37Z

Germans have a big advantage in those "Can you describe in one word...?" questions

I've been arguing for many years that the team is the real ...

2026-08-01T07:23:14Z

I've been arguing for many years that the team is the real product of software development, learning is the real outcome, and that organisations need to optimise for *that*

Software, and its impact in the real world, is an ephemeral byproduct of learning.

But teams aren't usually seen as an asset by management. When the software's shipped, orgs will often throw away the toy and play with the box.

You'd probably be a bit alarmed if your doctor had a Kanban ...

2026-07-29T08:31:58Z

You'd probably be a bit alarmed if your doctor had a Kanban board that showed you in the "Done" column just because the medicine had been dispensed.

There isn't a shred of hard evidence for any dev team getting ...

2026-07-28T10:30:57Z

There isn't a shred of hard evidence for any dev team getting 10x, 5x or even just 2x productivity using AI.

Software releases are questions, and only reality has the ...

2026-07-27T04:58:42Z

Software releases are questions, and only reality has the answers. Reality moves at its own pace and does not care about your ambition to 10x the process.

Ship 10x the software and you'll get 90% of those questions going unanswered until it's too late to change course.