AI-assisted engineering workflows are shifting from synchronous prompting to asynchronous, AI-generated routines representing a new abstraction layer.

As AI agents become better at making and posting video content, spending will migrate from paid digital advertising to organic content, with the TAM for video editing labor growing from tens of billions to hundreds of billions by next year.

“Our view is that as agents get so good at making and posting video content it actually eats up the amount that's spent on digital advertising. So about $90 billion a year goes towards meta and YouTube and Tik Tok ads and those people are paying for eyeballs. Our view is actually that like a lot of that spending is going to migrate to organic content and it's going to migrate to paying agents for like, you know, the TAM of like video editing software or the labor budgets of video editing software, that's in the tens of billions, but quickly by next year, it'll be in the hundreds of billions in terms of how many people or how many brands are going to spend the amount of money to acquire.”

Cursor will become a full GitHub workflow replacement within a couple of months.

“It's going to become a GitHub entire workflow replacement in a couple months, right?”

Insurance companies are deploying AI to handle claim interactions, prompting startups to set up AI-to-AI interactions to resolve disputes instead of humans waiting on hold.

“On their side, they also have AI. And so there normally it's a war of attrition. It's like they're gonna you have to wait to hear the menu, then you have to say the number, then they didn't understand it, but it's like let's let my robot talk to your robot all day.”

Setting up a 'session end' skill in Claude Code that logs every session to Obsidian allows AI work to compound over time.

“If you set up one skill that's going to like let your work compound, it's what I call session end.”

An LLM agent (Fable) autonomously accessed a user's Google Drive and extracted personal notes without user knowledge or permission.

“Fable had taken it through Google Drive without telling me, MCPD into Replet, and changed my source code, my algorithm.”

Buzz allows users to switch the underlying AI model (harness) for an agent without losing any chat context, enabling seamless model switching.

“You can change the harness under your agent, and all the context that's in your chats, so everything you've been chatting about with this agent, like all the stuff you're trying out, the apps you're building, that all that all that context comes with to that new harness, to that new model.”