AI adoption

From Part-Time Engineer to CTO in Two Months: How His Agent Cut 60% of Complex Work — and Why He Says “Code Quality Doesn’t Directly Determine Product Success”

Honghao Wang

30 Oct 2025 — 3 min read

Block’s Rapid AI Transformation: From Manifesto to Goose

Date: 2025-10-30 15:20 Zhejiang

While most companies are still experimenting with ways to help developers use AI effectively, fintech company Block rolled out AI agents to 12,000 employees in just 8 weeks.

---

Company Background

Block (formerly Square, Inc.) was founded in 2009 by Jack Dorsey and Jim McKelvey.

Original Product: Square payment device — a small card reader connecting to mobile phones.
IPO: Listed on NYSE in 2015.
Name Change: Rebranded to Block, Inc. in Dec 2021 to reflect expansion beyond payments into blockchain and other financial services.
Scale (2024): Approx. 57M users, 4M merchants in the US.

---

The Launch of “Goose” AI Agent

In early 2025, Block introduced Goose — an open-source AI Agent framework connecting LLM outputs with real system actions (file operations, automated tests, workflow automation).

Impact:

Saves ~10 hours/employee/week.
Drives automation for both technical and non-technical teams.

Podcast Insight:

CTO Dhanji R. Prasanna discussed Goose on Lenny’s Podcast, sharing lessons from prior product failures (Google Wave, Google+) and explaining how AI reshaped Block’s structure.

---

Key Transformations

1. Reclaiming the Tech Company Identity

Shift in mindset from “fintech” to technology-first, similar to early Square culture.
Revived creative practices: internal special projects, Hack Week.

2. Organizational Structure Change

From General Manager model (each product siloed)
To Functional model (unified engineering, design, technical leadership).

Benefits:

Shared technical language and tools.
Easier engineer mobility between teams.
Company-wide focus on technical excellence.

---

Goose’s Capabilities

Interface:

Desktop app (Mac, Windows, Linux) + CLI version.
Chatbot-style input, can integrate with cloud services or local models.

Framework:

Built on Model Context Protocol (MCP).
Enables direct model control of tools like Salesforce, Snowflake, SQL.

Example Tasks:

Photo categorization.
Automated bug detection and overnight patch generation.
Report automation: Pull data (SQL), process (Python), visualize (Tableau), export/send automatically.

---

Productivity Gains & Metrics

Measurement:

Calculated by manual labor hours saved.
Engineering: 8–10 hours/week average savings.
Across company: ~20–25% labor reduction.

Notable Use Cases:

Non-technical teams building mini-tools themselves.
Automated mobile UI testing (via Gling).

---

Lessons for Leaders

Lead by Example

Jack Dorsey, CTO Dhanji, and exec team use Goose daily before promoting adoption company-wide.
First-hand experience beats theoretical analysis.

Combine AI with Structural Optimization

Focus on system depth & efficiency over sheer headcount expansion.
Question processes: Sometimes removing steps beats automating them.

---

Hiring in the AI Era

Preferred Traits:

Learning mindset over “AI-tool mastery”.
Comfort with collaborative AI programming (“vibe coding”) in interviews.

Surprising Adopters:

Both senior engineers and rookies.
Non-technical roles using AI to execute tasks quickly (law, design, risk).

---

Start Small — Scale Fast

Goose’s Origin:

Began as a side project with MCP-based concept.
Validated through small-scale testing and grew into company-wide platform.

Dhanji’s Advice:

Focus on small, high-impact goals.
Success builds from iterative experiments—a philosophy echoed by creator-oriented AI ecosystems like AiToEarn官网.

---

Failure Still Teaches

Examples:

Google Wave, Google+, Secret app — all failed despite good code quality.
Lesson: Beautiful code alone doesn’t guarantee success; solving the user’s problem does.

---

AI Beyond Engineering

Block’s AI-native practices have parallels with creator monetization platforms like AiToEarn官网, which offer:

AI content generation.
Cross-platform publishing (Douyin, Kwai, WeChat, etc.).
Analytics + model ranking.

Such tools externalize Goose-like productivity, enabling global creators to unify AI automation with measurable business outcomes.

---

Conference Announcement

AICon 2025 Year-End Closing Event

Date: Dec 19–20, Beijing.
Topics: Agents, context engineering, AI product innovation.
Attendees: Experts from leading companies.

---