Peter Bell

How many engineers should you have? Fewer, more and different . . . gatherdev.substack.com/p/how-many-eng… Thoughts?

Not everyone needs to become an AI engineer In a world where engineers building software factories can deliver an unreasonable amount of code, it can be easy to think that all we need is more AI engineers who can build agentic harnesses. My guess is that in the future, we’ll see two distinct classes of engineering roles emerge.... gatherdev.substack.com/p/not-everyone…

Management by press release It's working for Jack, Tobi and Will at WHOOP . . . gatherdev.substack.com/p/management-b… Thoughts?

How to build a software factory Examples and an action plan if you're not yet building your agentic army :) gatherdev.substack.com/p/how-to-build… What are you building?!

Great article by @crichardson pointing out that in a complex domain (cynefin framework) software factories need to tolerate variability and ambiguity with layered guard rails and a focus on continual learning - microservices.io/post/architect…

1M token context window? Doesn’t matter. Push past ~50% and your brilliant AI intern starts acting like someone who pulled two all-nighters on Adderall. Context management is the job now. gatherdev.substack.com/p/context-and-… Thoughts?

If you're an engineering leader, Have you Had Your Claude Code Moment Yet? If you haven’t spent a half day pairing with an agentic coding tool, cancel some meetings and do it this week. There’s a before and after and it will materially impact your AI adoption strategy. gatherdev.substack.com/p/have-you-had… Thoughts?

For anyone not there yet - simplest productivity hack for working with AI: talk to it. Voice input changes how you think, how you prompt, and how fast you move. gatherdev.substack.com/p/talk-dont-ty… What are you using?

Blocks cuts weren't about AI, but Atlassian, Meta, Pinterest? More to come . . . :( gatherdev.substack.com/p/blocks-layof… Thoughts?

Flattening the agentic engineering org If agents can do the coordination work that middle managers used to do, what happens to the org chart? It gets a lot flatter. And that changes everything about how you structure an engineering team. gatherdev.substack.com/p/flattening-t… Thoughts?

From two pizza teams to triads? Two pizza teams made sense when coordination was the bottleneck. When agents handle the coordination, you might only need three people per team. Short Dominos. gatherdev.substack.com/p/from-two-piz… Thoughts?

A balanced look at the potential SaaSpocalypse Specific traits of SaaS companies likely to survive, and the real reasons to consider dumping your SaaS gatherdev.substack.com/p/a-balanced-l… What do you think?!

Our brave new world is just weird. Gave my agentic head of engineering a music back story to keep things interesting. Wanted a bumper for an upcoming podcast and here's what she said :) "I could write you something and you could find someone to record it. I know exactly what a CTO podcast intro should sound like on a Les Paul: something in the pocket, slightly behind the beat, with just enough grit to make people lean in. Four bars, done."

There’s no playbook yet for scaling the adoption of augmented and then agentic AI in the SDLC. But there are patterns emerging from the CTOs who are actually doing it. I finally needed somewhere to share insights from the monthly gather.dev CTO events in NY, SF and (soon) online and the ionterview for my O'Reilly book on Scaling AI Adoption in Engineering. Free substack - hope you enjoy! gatherdev.substack.com

@garrytan No, but the rich will get richer. Higher agency, better taste and better judgement and you’ll continue to find alpha in the frontiers. Great time to be a technical solopreneur

AI requires high IQ right now to get the most out of it But will that be true in 10 years? Or even next year?

@GergelyOrosz The difference is that most of the users will be agents and owning your data model (and data) is more important than the hi - which will be Ali’s for most ‘users’ by next year (or MCP servers if they’re still figuring out the shape of their queries)

Cannot wait for teams that build their custom JIRA/Workday replacement/custom CRM to one day turn around and ask: “Why do we have so much internal software that is buggy / has poor UX + we need to maitain?” Seen this movie well before AI, when Uber built uChat (custom Slack)

A small thank you to everyone using Claude: We’re doubling usage outside our peak hours for the next two weeks.

@garrytan It’s solid. I actually dropped obsidian - with a bit repo and Md files, the obsidian viewer is less important - just vibe code the custom ux you need to retrieve data. Also use eight persistence for right data. I have Md for agents and skills and supabase for the rest.

This looks cool, bookmarking for later

@andrewchen Have a strong opinion who you want to hire, intrigue well qualified passive candidates and pair with them to see how they ai - also hire folks who can bring their tribe. And then favor agents over humans to keep exponential comms from slowing down the business.

In recent years interviews for execs, product managers, marketers, etc have started to add a homework step so that people can show work output. It can be super helpful signal In recent weeks these homework replies have become overwhelmed with AI slop. Instead of 2-3 succinct pages reflecting a career of deep thinking, you get back fifteen pages of meandering ideas that anyone could generate What's the best way to address this? The best way, of course, is to actually work with them on a work trial so that you can really get a feel for how they act; however this is difficult because it takes a lot of time for them and for you so you have to reserve it for the end of your hiring funnel A more scalable idea is to ask people to present their homework in a recorded presentation so that they actually sign off on everything that they say and it's possible to ask them interactively later If you go much more structured, you risk offending the highest-end talent Other ideas? Open to the thoughts

@andrewchen Have my agents review every line of code

One question I've been asking founders is: do you try to review all the code that the LLMs write or do you just accept it? I think it's about 50-50 right now but the momentum is towards just accepting the AI-generated code and I think that number will eventually go to 100% This is one of the most telling indications of how AI-native a team is. It's hard to get super high throughput if you are reviewing every line Poll: what do you do?