What it means to show up

Hey neighbors,

A sad truth of showing up in the world is that, most of the time, you won’t get praise or recognition for putting yourself out there, but people do notice when you withdraw, or if you say you’ll show up and then flake.

I’d love to hear that someone noticed I skipped last Friday’s newsletter, that they missed hearing from me about stuff. But I also know that we all get a lot of newsletters, notifications, and other junk, and there’s a big difference between taking a week off and totally bailing.

A few years ago, I tried to make myself send a weekly update to my team at work, to start conversations and generally show people I was alive and paying attention. I made it two weeks, and the fact that I overpromised and failed to deliver came up later in peer review feedback. For my part, I had no idea anyone had noticed, because nobody piped up to say they had read the first two emails.

Showing up is hard if you don’t have a really clear idea of how to show up right, and sometimes you can only figure that out through trial and error.

Anyway, with this letter I’m trying two new things in addition to getting back on my Friday cadence. This note is around 400 words long (not 1,200 or 1,900), and it’s not about AI. Surprise! And you’re welcome! 🤣

New on the B&L Ideas blog

​No-code, vibe-code, same problems ​

Sorry, we’re back to AI. 😅 I’ve been seeing a lot of chatter on LinkedIn from marketers who’ll use Claude Code or similar tools to replace their Webflow sites with static Astro builds, seeing huge performance gains and generally feeling like superheroes. As I’ve seen more and more of these, though, I’ve concluded these are more lateral moves than upgrades, and these folks are likely to hit many of the same problems after the Claude honeymoon phase is over.

​Is it an agent, or just a cron job?​

Inspired by some recent client/prospect chats about “agents,” I set out to explain what makes something agentic as opposed to just regular old automation. And, rather than just make broad generalizations about performance, I ran some honest-to-god tests to find out the incremental cost of using AI for a scheduled job if you don’t actually need to.

More links to check out

​Design owns coherence — Peter Merholz shares some (as usual) excellent thoughts on the role of design in this crazy AI era. At Web Directions Product in Melbourne in 2019 — where I gave the opening keynote and Peter gave the closing one — Peter said something to the effect of, "product management is the new word for what we used to call user experience." By that he meant that, where companies had been looking to designers to make sure products were complete, correct, and fit for purpose, during the 2010s that role had been taken up by PMs in many companies.

Back in January Peter also wrote that product management has failed, or more accurately, that it's been collapsing on itself for lack of a clear definition or purpose. What's wild is that this lack of a clear, cogent role for PMs was the subject of my talk at that 2019 conference, The Product Polygon. At the time, no one was saying product had failed, and Peter's observation about PM being the new UX was a lot closer to right, though my point was that while PMs owned the experience, there was a lot of variance in how they did that.

Since COVID and ZIRP, the PM role has either shifted toward business, become all about writing AI evals, or (per Peter) collapsed entirely as a "glue" or "manager" role companies think they don't need.

Whether or not we agree that design necessarily owns coherence, someone has to, and products and companies do better when that person has a real seat at the table.

​The Feature Factory: A 10-Year Retrospective — Continuing on the "legacy of product management" theme, a decade ago John Cutler coined the term "feature factory" to describe tech product teams that were forced to crank out more and more stuff, as opposed to more enlightened ones that used discovery and product sense to deliver real value for users. Like Peter, John is a UXer turned PM, and his philosophy of product (like mine) is rooted in how teams can best serve users, as opposed to MBA-ish concerns like efficiency, throughput, and raw growth, and his reflections are a good summation of where tech is at in the present era.

​Tiny Little Office — From Paul Ford's AI-pilled agency/platform Aboard, this is a "full", if minimal, office suite with no accounts, no data storage, just documents and URLs. (Being such a browser-based, stateless app, that also means no AI, unless you object to the app itself being probably vibe-coded.)

​AI is a tool; using it well takes skill — Rich Hickey (creator of the Clojure programming language) recently argued AI is not a tool, because people use tools — directly, in their hands — whereas AI places distance between the user and the outcome. In response, Huahai Yang argues that AI is a tool because it requires skill, and the performance of an LLM depends on the quality of a user's instructions.

This is one of those situations in life where two things can be true. LLMs are not tools the same way screwdrivers or hammers are; to Rich's point, a drive-thru window is not a tool for getting fed. But good prompting does improve the work LLMs produce, which isn't to say the output will be proportionately as good as the prompt. Some excellent prompts will yield total bullshit.

Heeeeere’s Johnny

Today, John-John is pictured next to his spooky-season skeleton counterpart, questioning why I'm taking his picture instead of getting our walk on.