The mythical agent-month (News)

0:05Wes McKinney:What up nerds? I'm Jared and this is Changelog News for the week of Monday, February 23rd, 2026. Today we have some Changelog News news. This is my final episode. Starting next week, my good friend Adam Stachowiak will be taking care of you. Yes, after 13 years, 1 ,042 podcasts, 452 newsletters, and countless friends made along the way. It is time for a change. I will write more about this decision and the future on my blog,

0:39Andreas Kling:and we'll discuss it together on Friday's Friends episode, which will also be my last. Until then, thanks for logging with me, and definitely don't unsubscribe. I'm sure Adam's changelog news will be great.

0:51Wes McKinney:Okay, let's get in this week's news. The Mythical Agent Month. Wes McKinney has been wondering what many of us have been wondering. Quote,

1:15Andreas Kling:For now, Wes feels needed, but with things changing so rapidly, he wonders how much software engineering's past will inform software engineering's future. With that in mind, he decided to revisit one of his, and my, favorite books on the topic, Fred Brooks' The Mythical Man Month. In so doing, Wes discovered that the book's themes are still highly relevant in agentic software and that its follow-up, No Silver Bullet, predicts the exact problem he's having in his agentic engineering. Quote, The accidental complexity is no problem at all anymore, but what's left is the essential complexity, which was always the hard part.

1:51Andreas Kling:Agents can't reliably tell the difference. End quote. Thought-provoking stuff, definitely worth a read. Employ a peon today.

2:00Elliot Bonneville:I clicked the copy to clipboard button on the brew install command approximately 0.3 seconds after landing on this peon ping website. What could possibly be so compelling? Quote, game character voice lines the instant your AI agent finishes or needs permission. It works with Cloud Code, Codex, Cursor, OpenCode, Kiro, Windsurf, Antigravity, and more. never lose flow to a silent terminal again. End quote. With 95 plus sound packs and counting, there's a game character voice in here for everyone. I sampled a bunch, but I'm sticking with the default Warcraft 3 Orc peon for now.

2:38Wes McKinney:Ready to work.

2:39Elliot Bonneville:Give it a try. There is a 0 % chance this won't bring some joy to your work life.

2:44Wes McKinney:Me, not that kind of orc.

2:47Elliot Bonneville:Ladybird adopts Rust. One popular segment of our Ladybird pod with Andreas Kling and Chris Wanstroth was when Andreas told us they were leaning towards Swift as their C++ replacement. Well, well, well. How are the turntables? Quote, We previously explored Swift, but the C++ interop never quite got there and platform support outside the Apple ecosystem was limited. Rust is a different story. When we originally evaluated Rust back in 2024, we rejected it because it's not great at C++ style object-oriented programming. The web platform object model inherits a lot of 1990s OOP flavor with garbage collection, deep inheritance hierarchies, and so on.

3:27Elliot Bonneville:Rust's ownership model is not a natural fit for that. But after another year of treading water, it's time to make the pragmatic choice. Rust has the ecosystem and the safety guarantees we need. Both Firefox and Chromium have already begun introducing Rust into their codebases, and we think it's the right choice for Lady Bird 2. It's now time for sponsored news. What Spec-Driven Development Gets Wrong Spec-Driven development has a decay problem. Design docs go stale soon after they're written, and nobody gets rewarded for keeping them current. That was annoying before, and now it's getting dangerous.

4:02Elliot Bonneville:AI agents are following stale specs confidently, executing plans misaligned with assumed reality without ever flagging the drift. Here's what Amelia Wattenberger, product lead for intent from Augment Code, says in this post, quote, Every documentation-first initiative in software has failed for the same reason. It asked developers to do continuous maintenance work that nobody sees and nobody rewards. Augment's answer is bidirectional spec maintenance. Agents don't just read the spec, they write back to it. What happens when an agent discovers an existing auth context? It wires into that and updates the plan.

4:37Elliot Bonneville:If agents can write code, they can update the spec. This only works when the agent actually understands your entire code base. That's exactly what Augment's context engine is built to do. It opens the door for specs that get more accurate over time, not less. Learn more at augmentcode.com or follow the link to the full blog post in the companion newsletter.

4:57Wes McKinney:Cloudflare's new code mode technique. MCP is, for now at least, the standard way AI agents use external tools. But it sure does fill up the model's context window with a lot of cruft. To combat this, Cloudflare came up with code mode, which rhymes. so you know it's good. Quote,

5:15Andreas Kling:Code mode is a technique we first introduced for reducing context window usage during agent tool use. Instead of describing every operation as a separate tool, let the model write code against a typed SDK and execute the code safely in a dynamic worker loader. The code acts as a compact plan. The model can explore tool operations, compose multiple calls, and return just the data it needs. End quote.

5:39Elliot Bonneville:As a result of this, Cloudflare created a new MCP server for their entire API. Quote, with just two tools, search and execute,

5:48Andreas Kling:the server is able to provide access to the entire Cloudflare API over MCP while consuming only around 1 ,000 tokens. The footprint stays fixed no matter how many API endpoints exist. End quote.

6:00Wes McKinney:This is a good example of what I've been talking about a lot recently. The models are getting marginally better, but the traditional software engineering around the models squeeze out huge wins by equipping them better and better. The only moat left is money.

6:14Andreas Kling:Here's Elliot Bonneville. Quote, Every morning, a few thousand people wake up and ship something. A tool, a SaaS, a newsletter, an app that does the thing the other app does, but slightly differently. They post it on Hacker News. Nobody clicks.

6:27Elliot Bonneville:This is not new. What's new is the scale. An AI can wake up, or whatever it does at 3 a.m.,

6:33Andreas Kling:and ship 12 of these before breakfast. Creation used to be the scarce thing, the filter. Now attention is. Most of us are on the wrong side of that trade, end quote. Elliot says the effort it takes to build something is trending down, but the time we collectively have on this earth is fixed. In a world where attention is at a premium and slop abounds, attracting attention to your creation means you better have a headstart or a lot of money or both. Quote, the uncomfortable version. If you're not already moving, you might never take off. The cost of acting like this is true when it isn't is you move fast and spend money you didn't need to spend.

7:08Andreas Kling:The cost of acting like this isn't true when it is, that's permanent. End quote. Elliot's take is more doomery than I believe is warranted today, but I can't blame him. He shipped something new last week, and he's trying to attract attention to it. That's the news for now, but I'd like to take a moment to sincerely thank each and every one of you for reading, listening, submitting links, commenting, and just supporting my work all these years. It's been something special. Have yourself a great week. reach out and stay connected with me if you like my work and i'll talk to you again someday