Uploaded July 2026 | Updated November 2026, 5 days ago
Coinbase rebuilt its product development lifecycle around AI agents—and reduced the time from intent to production for mobile bug fixes and simple features from 20–25 days to approximately 1.8 days.
In this Arize:Observe 2026 session, Chintan Turakhia, Chief Builder and Head of Coinbase Wallet, explains what happened when his team paused shipping, deleted their IDEs, and redesigned product development for an agent-first world.
Instead of using AI only to generate code faster, Coinbase shifted human attention toward product intent, planning, validation, and risk. Chintan shares how two engineers rebuilt roughly 85% of the app in one week—and why the final 15% revealed the real challenges of AI-native software development.
The talk covers:
• Why Coinbase believes the traditional software development lifecycle is being replaced by a product development lifecycle
• How the team reduced the time from an idea to something users can test
• Why Coinbase paused shipping and asked engineers to delete their IDEs
• How AI coding agents exposed new bottlenecks in PR review, CI, builds, testing, and validation
• Why teams should evaluate the risk and health of an entire branch instead of reviewing every line of generated code
• How Coinbase removed standups, project kickoffs, PRDs, and unnecessary handoffs
• The “pirates and architects” model for building and scaling products
• How Forge turns Slack requests into PRs and one-off mobile builds
• How security agents review changes and assess deployment risk
• How Tracer Bullet workflows move from an idea to planning, code generation, simulator testing, verification, and internal release
• Why planning, context, and product visibility matter more as coding agents become increasingly autonomous
The core lesson: planning is the work. Give agents structured context and visibility into the product, then redesign the workflow around what humans should validate—not the code agents can already produce.
Chapters:
00:00 The SDLC is dead: Coinbase’s agent-first thesis
00:36 The stairway to AI-native engineering
02:38 Measuring time from intent to production
04:06 Pausing shipping and deleting every IDE
05:30 Rebuilding 85% of the app in one week
06:25 New bottlenecks: PRs, CI, builds, and validation
07:42 From code correctness to branch-level risk
09:45 Delete the process that no longer matters
11:53 Pirates, architects, and rapid prototyping
12:38 Coinbase’s agentic development stack
13:15 Forge: from a Slack request to a mobile build
15:52 Security agents and automated risk review
17:41 Tracer Bullet: taking an idea to production
19:27 Planning, execution, and agent verification
20:19 Plan, give agents eyes, and scale horizontally
21:25 Demo: parallel agents and device simulators
22:22 The one-day roadmap challenge
🔗 Learn more about Arize: arize.com
🔗 Explore Arize:Observe: arize.com/observe
🔔 Subscribe for more talks on AI agents, evaluation, observability, and AI engineering:
youtube.com/@arizeai?sub_confirmation=1
#AIAgents #AIEngineering #Coinbase
🔗 Try Arize AX & Phoenix OSS: arize.com
🔔 Subscribe for weekly content on LLMs, agents, and evaluation: youtube.com/@arizeai?sub_confirmation=1
Coinbase rebuilt its product development lifecycle around AI agents—and reduced the time from intent to production for mobile bug fixes and simple features from 20–25 days to approximately 1.8 days.
In this Arize:Observe 2026 session, Chintan Turakhia, Chief Builder and Head of Coinbase Wallet, explains what happened when his team paused shipping, deleted their IDEs, and redesigned product development for an agent-first world.
Instead of using AI only to generate code faster, Coinbase shifted human attention toward product intent, planning, validation, and risk. Chintan shares how two engineers rebuilt roughly 85% of the app in one week—and why the final 15% revealed the real challenges of AI-native software development.
The talk covers:
• Why Coinbase believes the traditional software development lifecycle is being replaced by a product development lifecycle
• How the team reduced the time from an idea to something users can test
• Why Coinbase paused shipping and asked engineers to delete their IDEs
• How AI coding agents exposed new bottlenecks in PR review, CI, builds, testing, and validation
• Why teams should evaluate the risk and health of an entire branch instead of reviewing every line of generated code
• How Coinbase removed standups, project kickoffs, PRDs, and unnecessary handoffs
• The “pirates and architects” model for building and scaling products
• How Forge turns Slack requests into PRs and one-off mobile builds
• How security agents review changes and assess deployment risk
• How Tracer Bullet workflows move from an idea to planning, code generation, simulator testing, verification, and internal release
• Why planning, context, and product visibility matter more as coding agents become increasingly autonomous
The core lesson: planning is the work. Give agents structured context and visibility into the product, then redesign the workflow around what humans should validate—not the code agents can already produce.
Chapters:
00:00 The SDLC is dead: Coinbase’s agent-first thesis
00:36 The stairway to AI-native engineering
02:38 Measuring time from intent to production
04:06 Pausing shipping and deleting every IDE
05:30 Rebuilding 85% of the app in one week
06:25 New bottlenecks: PRs, CI, builds, and validation
07:42 From code correctness to branch-level risk
09:45 Delete the process that no longer matters
11:53 Pirates, architects, and rapid prototyping
12:38 Coinbase’s agentic development stack
13:15 Forge: from a Slack request to a mobile build
15:52 Security agents and automated risk review
17:41 Tracer Bullet: taking an idea to production
19:27 Planning, execution, and agent verification
20:19 Plan, give agents eyes, and scale horizontally
21:25 Demo: parallel agents and device simulators
22:22 The one-day roadmap challenge
🔗 Learn more about Arize: arize.com
🔗 Explore Arize:Observe: arize.com/observe
🔔 Subscribe for more talks on AI agents, evaluation, observability, and AI engineering:
youtube.com/@arizeai?sub_confirmation=1
#AIAgents #AIEngineering #Coinbase
🔗 Try Arize AX & Phoenix OSS: arize.com
🔔 Subscribe for weekly content on LLMs, agents, and evaluation: youtube.com/@arizeai?sub_confirmation=1










