André Rocha (@andrerocha) profile picture

Product Management Executive, World traveler, Bicycle enthusiast, Guitar player,...

Sao Paulo, Brazil
Joined March 2007
Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)Photo by André Rocha (@andrerocha)
What's one thing you stopped doing yourself this year because an agent does it now, and one thing you took back?
3
My rule for agent-written code: a different model reviews it, in a fresh session. Same model, same blind spots. It costs a few minutes and catches what I'd miss at the end of a long day.
10
Promotions are a lagging indicator. You work at the next level for a while, then the title catches up. Asking for the title first is asking the company to bet on a version of you it hasn't seen yet.
1
34
Most experiments at small volume never reach significance. If the data shows it isn't making things worse and you believe in the direction, ship it and keep watching. Waiting for significance on a few hundred users is a slower way of not deciding.
24
André Rocha retweeted
Shoppers sent by AI show up ready to buy AI-referred shoppers vs. other online shoppers: - Conversion +60% - Revenue per visit +53% - Bounce rate -34% Charts of the Week: a16z.news/p/charts-of-the-we…
97
70
61
689
114,516
André Rocha retweeted
we will need to rebuild and rewrite much of software to adapt to a new, faster pace. code used to arrive one drop at a time. soon, if not already, code will be written as if it were a continuous stream, like a violent river twisting and turning as it pleases. this is not a new thing: big tech cos have had to deal with this with the enormous scale of their engineering teams. but now ai has brought an incredibly elastic and capable workforce wielded by every engineer. this is your problem now, and the only way to adapt is to evolve your systems so agents - even and especially the ones with little context about the right way of doing things - can thrive. make the easy thing the right thing. your codebase is becoming organic architecture. agent contributions grow around the constraints and guardrails you’ve designed. good software engineering matters even more now! and no, the continuous stream of code is not something you have to watch helplessly. be more ambitious and design systems, frameworks, and languages to control the chaos. the time for mourning coding is long over
Replying to @josephfcox
Here is Proton CTO's AI manifesto, that I obtained multiple copies of 404media.co/manifestos-fear-…
316
452
123
6,010
354,736
Edit for the audience, don't rewrite the strategy. If the same bet needs a different story for every room, you don't have a strategy yet. The claim should survive one telling.
1
38
I'd write this one-pager before giving any AI agent real work. No definition of done means it keeps going, confidently, in the wrong direction.
1
28
If an AI agent could take one recurring meeting off your calendar for good, which one? Mine would be the weekly status update. Curious if anyone would actually hand over a 1:1.
12
Roadmaps rarely fail because the plan was wrong. They fail because nobody wrote down what would make you kill it. Every big bet should ship with its kill criteria. If you can't name the number that would make you stop, you haven't decided yet.
1
10
Most teams adopt AI agents the way they adopted Slack: everyone at once, no rules. Six months later nobody knows which agent owns what. Mine each have one job, one owner, and one place to write. Boring structure beats clever prompts.
1
1
25
A feature nobody uses costs more than a bug. Bugs get fixed. Unused features get maintained forever, because nobody owns the decision to delete them.
16
What AI made cheap: First drafts, prototypes, summaries, code. What's still expensive: Deciding what not to build, saying no to a customer, earning a team's trust, owning the outcome. The bottleneck moved. It didn't disappear.
14
André Rocha retweeted
Asked my toddler for a good elliptic-curve private key and he said '4'
29
120
4
3,043
41,566
Product leaders: what's one decision you've fully handed to an AI agent at work, with no human review? Trying to figure out where the line actually sits for most teams.
28
Delegation fails less from bad people than from vague outcomes. 'Own onboarding' isn't delegation. 'Get more new users to their first win in week one, and you decide how' is.
18
André Rocha retweeted
ai agents will change your life bro. you go from spending the entire work day on your phone and computer to spending every waking hour on your phone and computer. from managing a team of people to managing a bunch of agents who are always up to some bs. i haven’t slept in a week
148
151
43
2,713
302,120
André Rocha retweeted
I think most people multi-tasking with agents could benefit from slowing down most of the stuff you're doing wont really matter- what's your P0? what would it look like if you gave it your full attention?
573
412
191
7,331
574,149
AI gave most PMs hours back every week. Very few spent them talking to customers. Most spent them writing more docs. The tool doesn't decide where the time goes. The team's habits do.
12
Crazy that a year ago we didn't even have skills
Today we're introducing Skills in claude dot ai, Claude Code, and the API. Skills let you package specialized knowledge into reusable capabilities that Claude loads on demand as agents tackle more complex tasks. Here's how they work and why they matter for the future of agents:
16