Writing

Claude isn't eating vertical apps. It's flooding them. Vertical SaaS startups are some of its biggest customers. Claude is incentivized to automate so much of the underlying workflow that every vertical is saturated with startups built on Claude. Voice agents for the trades was

Dinner w/ Harvard founders ('17-'25) last night, 43% use only Claude; 29% use only Cursor, rest on Claude + Codex. Hottest take at the table: strong IPOs will let labs overextend. Public market pressure will force them to increase cost of tokens, leading to a "correction".

Unique analogy from a founder: AI coding is stuck at Level 2 automation: impressive enough that devs stop checking, not reliable enough to stop checking. The same trap that plagued early self-driving. Self-driving is now at Level 4. What level are we striving for in software dev?

341K companies. $1.9T in IT spend managed. 1 Accenture. Others found a niche and never left. Picking the right wedge in AI IT services has never mattered more. https://t.co/pEOzRzU4z2

Enterprises spent a decade consolidating their data. Then they adopted a dozen AI tools in 12 months and fragmented it all over again. The rebundling has begun...and the frontrunners are selling unified platforms to the C-suite, not point solutions to departments.

From Pilots to Platforms: Picks for '26

Design at the Speed of AI

Cracks in the Checkout: Big Debates in Agentic Commerce

When Bots Outnumber Humans: The Case for Bot Tolls

Dear OpenAI, My Agent Needs Its Own Wallet

check out my new blog post here: https://t.co/MysBaVRim7

With 71% of Americans using rewards credit cards, and tools like Kudos and Google Wallet helping maximize points, these perks are bound to soar. But as merchants push back on the fees that fund these programs, how sustainable are these credit card rewards?