Clearly a bit biased, but good advice on feature flags that goes beyond Posthog.

I used a feature flags system like described here for something like a decade before moving to LaunchDarkly and it worked amazingly well. I’m not sure the move to LaunchDarkly actually paid off.

An interesting summary of expression end detection across a bunch of programming languages. Including Odin, which I have never heard of previously.

All of these lessons learned are ones I’ve seen before, but it’s a nice collection.

I particularly resonate with:

Admitting what you don’t know creates more safety than pretending you do.

and

Writing forces clarity. The fastest way to learn something better is to try teaching it.

An interesting use of nullcontext to have a parametrized test with valid and raising values.

An interesting categorisation of forks (towards the end of the post) from Monty of MySQL and MariaDB fame.

“Defense in Depth: A Practical Guide to Python Supply Chain Security”: an incredibly high quality post by someone with respectable credentials. Read the whole thing.

I like (a lot) webs of trust, but this human.json doesn’t seem like it will scale or give great information or really own the trust. I do like seeing example.com/~user style URLs, though, very nostalgic.

We need to expect to earn our keep. In the ZIRP era I think a lot of software devs lived in some world where they expected to make a lot of money in a way that was divorced from the realities of the business.

I feel there’s truth here, as much as I sympathise with those hit by change (and as much as it was painful when I was).

A depressing read on how ads have screwed up websites. I read almost everything either via RSS (NetNewsWire) or a read-later tool (Instapaper) so see almost none of this. The bit I hit most is the “CLS Disaster” on Goodreads, where I know now I need to wait when entering data because all the links I want will jump down the page after a giant ad loads at the top.

how come so many coders don’t just hate LLMs for stealing their work the way that most writers and photographers and musicians do? The answer boils down to three things:

Coders have long had a history of openly sharing code with each other, as part of an open source, collaborative culture that goes back for more than half a century. Tools for writing and creating code have almost always offered a certain degree of automation and reuse of work, so generating code doesn’t feel like as radical a departure from past practices. Software development is one of the fields with the least-advanced cultures around labor, as workers have almost no history of organizing, and many coders tend to side much more with management as they’ve been conditioned to think of themselves as “future founders” rather than being in solidarity with other workers.

Insightful comments on the likely impact of LLMs on dev work.

Handy suggestions for serving docs so that AI can get them more effectively. Unfortunately, with a ReadTheDocs site behind Cloudflare, I don’t think I can do any of this. Maybe some instructions to get the source files for specific sites, but they can’t be custom then.

Interesting information on the impact of current AI on employment, although from a very involved source. Somewhat terrifying that they have the data to do this, though.

“It’s like if the only way I knew how to make a hamburger bun was to carefully put every sesame seed by hand on the top only to stumble upon an 8 pack of buns for $4 at the grocery store after years of using tiny tweezers to put the seeds in exactly the right spot."

It doesn’t always feel like this, but it sure does sometimes. I’m sure there are people in the market for artisanal hand-crafted bun tops, and that it’s a great hobby for some, but it’s not going to hold up as a great career choice.

The level system (junior, mid, senior, staff) was never a development model. It was a compensation and expectations framework. It told you what someone should be able to do at each level. It said nothing about how they got there. But because the environment was producing growth on its own, the gap was easy to ignore. The labels tracked what was already happening, and everyone assumed the structure was the mechanism.

Interesting observations on the growth from junior to senior, and how AI changes that, or maybe for some managers doesn’t change it much at all.

“Never Send A Human To Do A Machine’s Job”. The first half is a great read on what the real benefit of code review is. And there there are some interesting thoughts on AI, which you can skip if you’re not into that, and still get much value from the post.

An interesting suggestion to start out by saying “first run the tests”. I’ll try to remember to give this a go a few times. It seems like it would be a bit annoying to do every single time. Although I guess an AGENTS.md could have it. It would put pressure on making running the tests much faster, because instead of being between the agent and getting the work done, it’s between me telling the agent what work to do.

This is very much a marketing post, but this:

Most code review tools start and end with the diff. They might expand to the file, or even the repository. But that’s not how the best reviewers think.

When a senior engineer reviews code, they’re drawing on years of institutional knowledge. They remember that Slack thread from six months ago about why the team chose this particular database pattern. They know that David on the platform team has strong opinions about error handling. They’ve internalized dozens of unwritten conventions that never made it into any style guide or lint.

is very true (it brings Chesterton’s Fence to mind) and not something that comes up in a lot of code review discussions.

these days I spend far more time thinking about how to make being wrong cheap rather than how to avoid being wrong.

Whenever someone used to ask me what my purpose was as an architect, this was my answer. It’s not about getting perfect designs, it’s about reducing regret. Maybe that is a great design that has flexibility and is so well made it doesn’t get changed – equally it could be something that can replaced with almost no regret.

A bunch of interesting thoughts on MCP vs. CLI and how it’s not “vs” at all. I’m not sure I agree 100% with this, but I do with a lot, and definitely the recommendation at the end. Follows on from this excellent earlier post.

The history of IDEs is so strange.

So true!

At some point frontier models will face diminishing returns, local models will catch up, and we will be done being beholden to frontier models. That will be a wonderful day, but until then, you will not know what models will be capable of unless you use the best. Pay through the nose for Opus or GPT-7.9-xhigh-with-cheese. Don’t worry, it’s only for a few years.

Sad, but true.

use a fresh VM

There have been done interesting UI changes here recently, with things backing that underneath I assume, but still entirely true.

Read the whole thing.

AI.

Sigh.

They rhyme for a reason.

Read the whole post.

Interesting suggestions for security policies, aimed at AI reports, but generally useful in today’s world.

Now, before some of you cry “but, but it’s not production quality“ or “non-developers don’t know what they are doing, they will hire a developer to fix their vibe coded mess”.. here’s what many don’t get: Businesses run on a ton of “brittle spreadsheets” already. People have been “vibe spreadsheeting” for a long time.

A good insight. In-house software is (especially outside of tech organisations), in my experience, terrible. Awful UX, buggy, half-done. And continually used for critical functions.

When a colleague writes code, I know their patterns, their strengths, their blind spots. I can skim the parts I trust and focus on the parts I don’t. With AI, every line is suspect.

Siddhant Khare on AI Fatigue. I 100% agree with the above, and I think I’m feeling some of this, too.