Work with me
I build systems where the machine proposes and a human decides.
AI agents today, physical AI next. Alongside my job as CTO at Bestie Bite, I take on a few teams at a time: dev teams that want agents with a person in charge, and robotics startups whose robots work outdoors alongside people.
AI agents
AI agents that do the work. You keep the last word.
In 4 weeks your team gets agents that review pull requests, pull commitments out of meetings and propose decisions. You approve from your phone in about 15 minutes a day. Nothing is merged or decided without a person’s tap.
Who it’s for
CTOs and VPs of Engineering at seed to Series A startups with a dev team of 3 to 15 people, already using Claude Code or Copilot, in Italy or the EU. It is not for teams that want agents to decide on their own.
How it works
The agent proposes. A human approves with one button. I built this method and measured it on a real team of three:
- 54 pull requests reviewed by agents in one week. A person merged every one.
- 307 decisions proposed in one week, 149 answered.
- 26 meetings, 158 commitments extracted, 6 proposed, 4 applied.
- ~40 agent jobs on a ~€55/month VM plus a Claude subscription.
The four weeks
- Days 1–2
- Setup, remote is fine. We pick 3 jobs, then I set up your team’s CLAUDE.md, skills, hooks and guardrails, the approval rule, an undo for every applied change, a notification gate with quiet hours, and cost tracking from day one.
- Week 1
- The pull request review agent goes live and we tune it on your codebase.
- Week 2
- Meetings become commitments, commitments become proposals.
- Week 3
- Decision proposals, undo and the notification gate.
- Week 4
- We measure approvals per day, the minutes you spend and the running cost, and I hand over your team’s playbook.
During the 4 weeks: written answers within one working day and a 30-minute check-in every week.
What you get
- A working approval system in your repos, with 3 jobs running.
- Your team’s CLAUDE.md, skills, hooks and guardrails, written for your codebase.
- Undo, a notification gate, quiet hours and cost tracking.
- A short playbook, so your team can add the next jobs without me.
- Price
- From €6,000, fixed after the Approval Audit. Optional monthly support afterwards, 4 to 6 hours a month. I take at most 2 teams per quarter.
- Guarantee
- If after 4 weeks your team isn’t using the system every working day with human approval, you get one more month of support, free.
Book a free 30-minute Approval Audit →
We pick the 3 jobs to automate first in your team. Two slots a month. Prefer writing? Answer five questions by email and get a written reply within 48 hours.
Robotics
Advisor for startups building robots that work alongside people, outdoors.
I help robotics teams take robots out of the lab and into fields, vineyards, orchards and farms, where they work next to the people picking, pruning and carrying. My PhD at the University of Trento was on predicting how people move and on shared control for robotic walkers. Since then I’ve been turning technology into a product as a CTO.
Where I help
- Next to people
- Human-aware navigation and human motion prediction, so the robot is legible and the people working next to it in the field accept it.
- Shared control
- Shared authority between robot and person (picker, operator, teleoperator): who decides what, when the robot proposes and when it hands over control.
- Pilot to service
- What to keep from the research and what to rebuild, the architecture, the first engineers, and a technical roadmap investors understand.
- Advisor
- Equity only, 0.25–0.5%, 2–3 hours a month plus async. At most three roles at a time.
- Review
- Not at equity stage yet? A paid 2-week Robotics Readiness Review (€4–6k): how your robot moves around the people working next to it and shares control with them, and what is missing to go from pilot to paid service.
- Not for
- Food or food-tech startups.
Or send me two lines on what you’re building and where you’re stuck: hello@placidofalqueto.com.
Questions
- Do I get your company’s code?
- No. You get the method and a setup written for your repos. I don’t share code or internals from the company where I’m CTO.
- What if an agent gets something wrong?
- Agents propose; a person approves. Every applied change has an undo, and nothing is merged without a human.
- What does it cost to run?
- Our reference: about 40 jobs on a ~€55/month VM plus a Claude subscription. Yours depends on the jobs; you see the cost from the first day.
- We use Copilot, not Claude Code. Does it still work?
- The agents run on Claude Code. Copilot can stay in your developers’ editors.
- Can it all be remote?
- Yes. The 2 setup days can be remote, and split over two weeks if that suits your team better.
- Why only 2 teams per quarter?
- I’m a full-time CTO. Two teams is what I can do well.