AI coding agents will not kill software engineering.

a 🧵 ...

Friday mini-rant: Talking to engineers, leaders, founders... I'm hearing such a big gap between how these different groups feel about AI tools and productivity.

Pure excitement today! We have an MCP server in beta that helps ground coding agent in context, best practices, etc...

We're seeing software experimentation move up the stack.

Once the "what to build" is well-defined, lines of code can be written extremely fast by AI-enabled engineers.

Interesting observation - in the software space, details are getting further away from more people. That means the risk of bad business outcomes due to invisible details is growing.

The last time I posted about LLM data use policies was... before DeepSeek so... here's an update! 😂

We've been on quite the ride with AI products:
- ChatGPT launches
- "Wrapper" products flood the market
- Foundation model companies release features that kill wrappers
- DeepSeek announcements drive AI costs down
- We're back to... wrappers?

In December, for the first time ever, officials from the US government were recommending that the public use end-to-end encryption for normal communications (!!!!!) yet somehow the story flew under the radar.

Context is everything (and it always has been).

A lot of people get frustrated that AI can’t solve all their problems, but useful AI-based solutions can still be built.

Just your friendly reminder to bounce that thought off a friend ✨

I've easily spent over 1000 hours coding with various AI-based tools. They're not replacing software engineers who are great at value-based and systems thinking.

More news, more reminders that privacy isn't built-in to every platform...

Encryption is back in the headlines, and the FBI and CISA are recommending encrypted messaging services. What do we need to know?