About Dani Prasana
Dani Prasana is the CTO of Block (Square, Cash App, Afterpay, Tidal), overseeing 3,500 people. He pioneered AI-native development at Block with Goose, an open-source AI agent that he has watching his screen all day.
Career Highlights
- Block (present): CTO, overseeing Square, Cash App, Afterpay, Tidal engineering
- Goose: Built open-source AI agent (any model, MCP integrations)
- GM to Functional reorg: Consolidated all engineers under one team for AI transformation
- Jack Dorsey letter: Wrote manifesto that catalyzed Block’s AI shift
Notable Positions
On Measured AI Productivity
Real metrics, not hype:
“8-10 hours saved per week - self-reported, then verified with PR throughput and a complicated formula our data scientists built. Across the whole company, we’re trending toward 20-25% manual hours saved.”
On Screen-Watching AI
The most aggressive AI adoption pattern:
“One engineer has Goose watching his entire screen. He’ll discuss a feature on Slack and hours later Goose has already tried to build it and opened a PR. It pushes him out of meetings that go too long. You have to have the courage to be that tightly coupled to your tools.”
On Non-Engineers Building Software
The biggest surprise:
“Enterprise risk management built their own self-service system. Finance, legal - building software tools that compress weeks into hours. No waiting for Q2 planning from internal application teams.”
On Conway’s Law
Organizational change required:
“Jack Dorsey came back, 40 execs in weekly meetings, but nobody talked about AI. I wrote a letter. Then we reorganized from GM silos to functional organization - all engineers under one team. We’re building to our org structure.”
Key Quotes
- “8-10 hours saved per week, verified.”
- “Goose watches my screen all day.”
- “Non-engineering teams have the biggest impact.”
Related Reading
- Enterprise AI - Block’s transformation story
- Application Over Training - The product focus Block embodies
Video Mentions
AI productivity metrics
8-10 hours saved per week - self-reported, then verified with PR throughput metrics. Across the whole company, support, legal, risk - not just engineering - we're trending toward 20-25% manual hours saved.
Screen-watching AI
One engineer has Goose watching his entire screen. He'll discuss a feature on Slack and hours later Goose has already tried to build it and opened a PR. It does things he didn't ask for but that it thinks will be useful.
Non-engineers building software
Biggest surprise: non-engineering teams. Enterprise risk management built their own self-service system. Finance, legal - building software tools that compress weeks into hours. No waiting for Q2 planning from internal application teams.