Everyone feels behind
Every company needs to challenge how it works with AI. Provocations from Spawn, Mark Cuban, and Martin Gratzer, plus agent security and engineering interviews.
You are not behind. You are absolutely behind. Everyone is behind.
It’s easy to feel like all of these things are true. And maybe they are! It’s not entirely clear. The truth is that every company is different. There are companies out there that are almost certainly moving faster than you. There are also plenty of companies out there just getting started. And truly nobody knows what they’re doing (everyone at the edge is making it up as they go along).
This post from Nichole Wischoff gets at what lots of people are feeling:
Scrolling X for five minutes: Omg someone is building everything and has raised so much money, market is taken. It’s too late.
Talking to any large private or public company CEO: Uh, we don’t use AI, working on a strategy, also how do we price agents?
Most likely you are somewhere in between those two, which means you’re in the big messy middle along with almost everyone else. It’s okay. Take a deep breath. And then get to work.
- Feeling behind is a great motivator. It’s a simple truth that you can be doing more.
- You will not stop feeling behind by being conservative, or by doing things the way everyone has gotten comfortable doing them.
- You do not need to operate like Anthropic overnight, but you also need to stop operating like you did 12-24 months ago.
- Choose a small experiment to run with your operating model. One week, or a few, of doing things differently. Permission to flout the norms (at least the ones that won’t get you in regulatory trouble). Take stock of what worked and what didn’t; rinse and repeat.
The important thing is not where you’re at today. It’s how willing you are to adapt and change.
Here’s what we’re reading this week:
How I Built the Most AI-Pilled Company in the World
Jacob Sansbury, founder of Spawn. The post is a bit braggadocious and will absolutely not apply to every org (or even most orgs). But we think it’s worth keeping tabs on what people are doing at the edge, and Spawn definitely seems to be at the edge. Some interesting notes:
- They take a hard stance that Anthropic models are the only ones currently capable of supporting their highly autonomous workflow.
- They (mostly) merge to main to avoid PRs and CI as a bottleneck, reconciling problems on main and prioritizing quick fixes to production.
- They don’t use the word “factory” once, instead describing a system where work is initiated, built, and deployed to production that bears only a passing resemblance to how software was developed everywhere a few years ago.
The second [job a human has in this era] is entropy, asking the models for things they wouldn’t have suggested on their own. Ambition is one kind. Creative entropy is another.
Mark Cuban on AI adoption and the opportunity ahead
That’s the brilliance of AI, right? AI is not smart. It’s book smart maybe, but it’s not street smart. It’s not entrepreneur smart. You’ve got to have somebody there who guides it and helps it.
Mark Cuban on why it’s important for entrepreneurs and C-suite leaders to experiment and understand what’s possible with AI.
Reforging the Workflow: Where the Human Goes When Agents Ship
Martin Gratzer, Staff Engineer at Meister, follows up on a post from February to revisit how he’s using AI now compared to 8 months ago. It’s a low-hype, level-headed, experience-based analysis. In February he was convinced that human review was crucial. Now he challenges that conviction and describes where he’s shifted investment instead.
Don’t get me wrong, I’m still in the loop and I’m still in control. I just spend less time reading diffs and more time at both ends: defining the problem before the agent starts, and shaping the system it works in.
Replit’s Architecture Behind Secure Agent Work
Given the spate of recent cybersecurity issues, security is a topic on many leaders’ minds. This Replit post details how they prioritize security in an agentic world through access controls, containing failures, and governance. It’s a good read for anyone who cares about security (which should be everyone).
Hunger beats tenure: Advice for junior engineers
Felix Rieseberg from the Claude team is back to declare the death of the coding interview and provide some practical advice for early-career engineers. If teams at the edge like Anthropic are questioning the value of the coding interview, it’s probably worth thinking about what that may mean for your team, if not now then in 3, 6, or 12 months. And Felix lays out both the advantages and disadvantages that less experienced engineers have today, and how they can make the most of the opportunity (or how you, as a leader, might ensure those opportunities are within reach).
Felix relays the story of how the Claude Code desktop app was born:
At Anthropic, people got so used to using Claude Code as a command-line interface (CLI) that it took two relatively new engineers to whip up a working prototype over one weekend. The first public version shipped roughly eight days later.
What we’re talking about this week
Testing every personal AI agent
Claire created a personal agent benchmark and tested them all so you don’t have to. One came out on top as the clear winner.
Claire has been exploring using agents for video generation and turned some content we had written about decision models into a 3-minute high-quality video explainer. We’re biased, but...it’s pretty good.
One thing to try this week
We’ll be giving Claude and Codex a weekend respite so we can spend a spooky Halloween Saturday with our respective little trick-or-treaters. Dress up, have fun. AI will be here when you get back.
Happy Halloween!
— Claire + Zach