|
AI Builders Digest
Friday, July 17, 2026
|
|
Box CEO Aaron Levie has now been in rooms with both developers and enterprise IT leaders in the span of 48 hours, and the gap between those two conversations is the most useful thing in today's feed. Developers are shipping. IT leaders are still figuring out where to put their data.
|
|
---
|
|
01
|
Enterprise agents are stalling on a problem that has nothing to do with AI
|
|
|
Levie hosted a dinner with IT leaders from large companies this week and came away with a clear pattern: change management, not model capability, is the actual bottleneck. Most enterprise processes weren't designed to hand off decisions to an agent. The data is scattered, the workflows are manual, and the humans in the loop haven't been told what "working with agents" actually means for their jobs. Levie noted that IT teams are spending significant energy just getting structured and unstructured data into a state where agents can use it at all.
|
Why it matters: If you're a vendor selling agent software to enterprises, your real competition isn't another AI company. It's the IT leader who has a two-year data cleanup project ahead of them before your product can do anything useful. And if you're inside a large company planning an agent rollout, the org chart and the data architecture need to change before the model does.
|
|
Source →
|
|
---
|
|
02
|
Google's Gemini Spark gets faster and learns to edit your Docs
|
|
|
Josh Woodward, who leads Gemini at Google, announced four updates to Gemini Spark rolling out to more Ultra subscribers globally. The headline changes: Spark can now open and edit Google Docs directly, read comments in Sheets and Slides, and Google claims it's more than 50% faster than the previous version. It can also work across multiple sources at the same time rather than sequentially. Pro tier access is coming separately.
|
Why it matters: The Docs editing capability is the one to watch. Once an AI agent can read your comment thread and then rewrite the document, the "just use Google Docs" workflow gets meaningfully different. Your team's review process, which currently runs through comments and track changes, now has a new participant who acts on it automatically.
|
|
Source →
|
|
---
|
|
03
|
Vercel's analytics API is becoming an agent data layer
|
|
|
Following yesterday's coverage of Vercel CEO Guillermo Rauch's AgentMail push, Rauch is back with another piece of the same puzzle. He posted use cases for Vercel's Web Analytics API where agents correlate site traffic, custom events like purchases and checkouts, with deployment timelines and performance data, then plot that alongside Stripe and Resend data for a unified view.
|
Why it matters: Rauch is quietly assembling something: a platform where agents don't just deploy code but also read the business signals that code generates. The vision is an agent that ships a feature, watches what it does to conversion, and connects those dots without a human pulling reports. Whether that lands or stays a demo-circuit idea is the open question.
|
|
Source →
|
|
---
|
|
|
|
**Coding agents as a creative medium, not just a productivity tool** — Zara Zhang, who works in AI investing, posted that because she never learned traditional programming, using coding agents has been "purely an act of creativity and self-expression." Her GitHub is her Substack. Worth sitting with: the first generation of people who build software entirely through agents may think about it more like writing than engineering.
|
|
Source →
|
|
**Swyx's post has no substance worth covering** — "You can just tweet things into existence" with a screenshot link tells us nothing we can pass on. Skipping it in good conscience.
|
|
Source →
|
|
Follow builders, not influencers. A daily digest of what matters in AI.
Read online ·
Archive
|