
Eng Leader · Co-founder @cxodev
Oakland, CA
Joined February 2008
- Tweets277
- Following104
- Followers1.4K
- Likes85
zachdavis retweeted
I’m not an eng, but have grappled with this on our team:
1) work at the fidelity most likely to accelerate velocity, go lo fi for high level decisions (ie don’t default to hi fi prototypes, be deliberate)
2) generate more, but spend most time on subtraction
zachdavis retweeted
In design interviewing (in my past life), there was a notion of “Design rationale” - seeing if the candidates were able to articulate design decisions clearly.
Been seeing the evolution of prototyping, and now it is technically easier to build (design+code) something.
Designs that actually ended up shipping especially in large team settings were the ones that had 1) clear creative intent, 2) business context, 3) steer the stakeholders towards consensus.
It’s probably a new era that just because it’s real code doesn’t mean it’ll ship. Creative vision is hard!
I've talked to multiple design teams in the last week worried about fidelity of AI-generated prototypes conveying a sense of completeness or depth of thought that can be harmful.
This is obviously a problem we've seen before (and continue to see), both in generated prose and generated code. What have we learned from other domains that we can bring over to our design friends?
I've talked to multiple design teams in the last week worried about fidelity of AI-generated prototypes conveying a sense of completeness or depth of thought that can be harmful.
This is obviously a problem we've seen before (and continue to see), both in generated prose and generated code. What have we learned from other domains that we can bring over to our design friends?
With code generation I think we've mostly gotten to a point where it's a moot point, because the models produce better code and because verifiable correctness often outweighs craft or taste.
Prose and design don't work that way yet, and may never work that way.
2
5
1,280
So what's the answer for design? I'm still not sure, but I will offer this half-baked idea to provoke thought:
Maybe the answer is to be more ambitious? This is what actually happened in eng IMO. There was (and is) lots of hand-wringing but if you lean into the tools and the possibilities you realize you can actually do way more at a quality that previously seemed impossible.
So maybe it's likewise possible to regularly create high-fidelity, well-considered mockups or prototypes that can be shared and built if there's consensus, but you also don't mind throwing away (I don't know, I'm not a designer, I just sometimes do design-y things)
2
9
1,126
I know I'm late to the factory fight, but I've been thinking about whether and how to apply (the best of) factory concepts to other (non-software) work.
I think most people mistake automation as the key characteristic of a factory.
A factory isn't about what it produces, or how automated it is. It's about systematically measuring, controlling, and improving a production process.
1
6
300
Shout out to @zachlloydtweets who's writing about software factories is really clear-eyed and something I kept going back to as I was thinking about how to factory-ify other work.
1
286
zachdavis retweeted
This is SO much fun. Love hearing @clairevo and @zachdavis break it down; even if it is their robot dopplegangers.
Such a great format.
I like the take on verification here. The way we "check" code is changing A LOT. Verification is the way forward 🙏
INTRODUCING: 🎧 Agent-Ready with @cxodev
Every week, @zachdavis and I share
→ what we see inside AI-pilled teams
→ the releases + products to pay attention to
→ who's inspiring us to think bigger
Written by humans, voiced by AI, with stories illustrated by our team of CXO.dev bots.
Trying to move faster with AI won't fix your org's culture and operating model problems. It will expose them and make them more painful.
If you're not willing to invest in fixing those problems you're going to have a bad time.
4
666
One thing I really appreciate about Grok Bot in Slack is that it's well tuned to not respond to every single message in a thread. Often there are side chats within a thread and previous generations of always-on agents would over-eagerly respond to everything, which was very annoying. Grok Bot seems to know when it's needed (and when it's not).
1
3
534
This is a good list. But whether or not you agree that these 5 are the right things to focus on, once you have a list like this it doubles as a list of the things you should optimize and focus on to ensure changes move quickly and safely through your system ("factory" or not).
Once the agent has reviewed and verified its own work, the human reviewer needs to understand the scope of the change, and which parts are worth the closest review. These are five criteria worth assessing:
1. Verification: Is the code well verified, both with test coverage and end-to-end computer use artifacts?
2. Alignment: Does the code follow the issue and PR description, or are there drive-by fixes and unintended side effects?
3. Impact: What is the blast radius of the change, and is it touching sensitive code paths like auth, permissions, or billing? For sensitive systems, this is where you should route reviews by code owner so the right person is tagged into the PR.
4. Rollback cost: How easily can this be rolled back, and what is the rollback procedure in the case of an incident? Matt Pocock frames this as a one-way door versus a two-way door, where a one-way door (like a destructive schema migration) is harder to reverse.
5. Error detection: If something goes wrong in production, how much visibility will we have? Are errors swallowed, logged, or reported to an error monitoring service? That last option is ideal for agent visibility to fix their own mistakes. For example, we added a webhook to monitor for Sentry alerts and relay reports to our software factory agents for investigation.
These five categories help you decide how much human oversight a PR actually needs.
2
415
Are your engineers ripping PRs, but everything's getting stuck in code review? 🫠
You're not alone.
📅 Thursday, November 19
→ Review is the New Bottleneck: Rethinking Code Review
→ @clairevo and @zachdavis talk about how to speed up shipping, without breaking your code (or your team)
→ Presented by @Amplitude_HQ
Registration now open: cxo.dev/events/review-is-the…






