Tyler
@tylerfolkman.bsky.social
OpenAI “open-sourced” their shopping protocol but kept the important part closed: product discovery and ranking.
You can build payments with it. But they control which products 700M users actually see.
It’s not open commerce. It’s a walled garden with open plumbing.
You can build payments with it. But they control which products 700M users actually see.
It’s not open commerce. It’s a walled garden with open plumbing.
October 2, 2025 at 2:03 PM
OpenAI “open-sourced” their shopping protocol but kept the important part closed: product discovery and ranking.
You can build payments with it. But they control which products 700M users actually see.
It’s not open commerce. It’s a walled garden with open plumbing.
You can build payments with it. But they control which products 700M users actually see.
It’s not open commerce. It’s a walled garden with open plumbing.
Your AI gets dumber with every token you add.
LLMs have an "attention budget." Just like you forget everything from meetings over 30 minutes, Claude forgets the middle of your massive prompts. Anthropic calls it "context rot."
LLMs have an "attention budget." Just like you forget everything from meetings over 30 minutes, Claude forgets the middle of your massive prompts. Anthropic calls it "context rot."
October 1, 2025 at 3:29 PM
Your AI gets dumber with every token you add.
LLMs have an "attention budget." Just like you forget everything from meetings over 30 minutes, Claude forgets the middle of your massive prompts. Anthropic calls it "context rot."
LLMs have an "attention budget." Just like you forget everything from meetings over 30 minutes, Claude forgets the middle of your massive prompts. Anthropic calls it "context rot."
Your AI coding agent is programming with a blindfold on.
Their new DevTools MCP server lets AI agents debug live in Chrome. They can now:
- See console errors in real-time
- Diagnose why images won’t load
- Run performance traces
- Inspect the DOM like a human would
Their new DevTools MCP server lets AI agents debug live in Chrome. They can now:
- See console errors in real-time
- Diagnose why images won’t load
- Run performance traces
- Inspect the DOM like a human would
September 30, 2025 at 2:19 PM
Your AI coding agent is programming with a blindfold on.
Their new DevTools MCP server lets AI agents debug live in Chrome. They can now:
- See console errors in real-time
- Diagnose why images won’t load
- Run performance traces
- Inspect the DOM like a human would
Their new DevTools MCP server lets AI agents debug live in Chrome. They can now:
- See console errors in real-time
- Diagnose why images won’t load
- Run performance traces
- Inspect the DOM like a human would
Claude Code 2.0 just dropped 2 hours ago.
Checkpoints let you rewind any AI code changes (Esc twice).
Subagents running parallel workflows.
VS Code extension in beta. Real-time diffs in a sidebar.
Anyone else eyeing this for their team?
Checkpoints let you rewind any AI code changes (Esc twice).
Subagents running parallel workflows.
VS Code extension in beta. Real-time diffs in a sidebar.
Anyone else eyeing this for their team?
September 29, 2025 at 6:54 PM
Claude Code 2.0 just dropped 2 hours ago.
Checkpoints let you rewind any AI code changes (Esc twice).
Subagents running parallel workflows.
VS Code extension in beta. Real-time diffs in a sidebar.
Anyone else eyeing this for their team?
Checkpoints let you rewind any AI code changes (Esc twice).
Subagents running parallel workflows.
VS Code extension in beta. Real-time diffs in a sidebar.
Anyone else eyeing this for their team?
Stop arguing about AI productivity. Start measuring it.
Most dev teams debate whether AI makes them faster while flying completely blind on actual metrics.
If you’re a CTO or engineering leader running teams using AI tools, this is the playbook.
open.substack.com/pub/tylerfol...
Most dev teams debate whether AI makes them faster while flying completely blind on actual metrics.
If you’re a CTO or engineering leader running teams using AI tools, this is the playbook.
open.substack.com/pub/tylerfol...
How to Actually Measure AI's Impact on Your Dev Team
Is AI making you faster or just making bigger PRs? Are you shipping quality or just feeling productive? Here’s a 2-hour setup to get real data on your AI productivity.
open.substack.com
September 28, 2025 at 5:54 PM
Stop arguing about AI productivity. Start measuring it.
Most dev teams debate whether AI makes them faster while flying completely blind on actual metrics.
If you’re a CTO or engineering leader running teams using AI tools, this is the playbook.
open.substack.com/pub/tylerfol...
Most dev teams debate whether AI makes them faster while flying completely blind on actual metrics.
If you’re a CTO or engineering leader running teams using AI tools, this is the playbook.
open.substack.com/pub/tylerfol...
GitHub released Copilot CLI in public preview. This isn’t innovation anymore. It’s validation that terminal AI is now table stakes.
Six months ago Claude Code was an experiment. Today every serious developer tool needs a CLI with conversational AI.
Has anyone tried it yet? Planning on using it?
Six months ago Claude Code was an experiment. Today every serious developer tool needs a CLI with conversational AI.
Has anyone tried it yet? Planning on using it?
September 27, 2025 at 8:59 PM
GitHub released Copilot CLI in public preview. This isn’t innovation anymore. It’s validation that terminal AI is now table stakes.
Six months ago Claude Code was an experiment. Today every serious developer tool needs a CLI with conversational AI.
Has anyone tried it yet? Planning on using it?
Six months ago Claude Code was an experiment. Today every serious developer tool needs a CLI with conversational AI.
Has anyone tried it yet? Planning on using it?
60% of engineering leaders can’t measure if their AI tools actually work.
Here’s what Google, Microsoft, and 16 other top companies actually track:
• Change failure rate paired with PR throughput
• Developer satisfaction scores
• Time to ship features customers pay for
Here’s what Google, Microsoft, and 16 other top companies actually track:
• Change failure rate paired with PR throughput
• Developer satisfaction scores
• Time to ship features customers pay for
How tech companies measure the impact of AI on software development
How do GitHub, Google, Dropbox, Monzo, Atlassian, and 13 other companies know how well AI tools work for devs? A deepdive sharing exclusive details, with CTO Laura Tacho
newsletter.pragmaticengineer.com
September 24, 2025 at 9:59 PM
60% of engineering leaders can’t measure if their AI tools actually work.
Here’s what Google, Microsoft, and 16 other top companies actually track:
• Change failure rate paired with PR throughput
• Developer satisfaction scores
• Time to ship features customers pay for
Here’s what Google, Microsoft, and 16 other top companies actually track:
• Change failure rate paired with PR throughput
• Developer satisfaction scores
• Time to ship features customers pay for