|
AI Builders Digest
Thursday, July 16, 2026
|
|
The theme today is builders finding creative uses for the tools in front of them, from Pokémon team analysis to cryptographic proofs. But the most telling item is OpenAI handing out $100 in credits to anyone who tweets something nice about their model. When your go-to-market strategy is a testimonial campaign, you're in a different kind of competition.
|
|
---
|
|
01
|
OpenAI is paying people to say they switched to GPT-5.6 Sol
|
|
|
Thibault Sottiaux, who works on Codex at OpenAI, posted an offer: tweet what you love about GPT-5.6 Sol or explain why you switched to it, and OpenAI will send you $100 in Codex credits. First 10,000 people. The replies are currently a mix of genuine enthusiasm and people who clearly just want the free tokens.
|
Why it matters: This is a testimonial campaign for a coding AI product, which means OpenAI thinks the battle for developer loyalty in this space is close enough to contest with incentives. Yesterday we noted Swyx's multi-model workflow where developers are mixing Claude and GPT for different tasks. A $100 bounty for a pro-GPT tweet suggests OpenAI is watching that behavior and doesn't love it.
|
|
Source →
|
|
---
|
|
02
|
Vercel opens its AI traffic data to the public
|
|
|
Guillermo Rauch, CEO of Vercel, posted that the company is releasing the dataset of AI token flows from the Vercel AI Gateway. This is real usage data showing which models developers are actually calling, in what volumes, and in what patterns.
|
Why it matters: This is the most honest picture of what developers are actually building with AI, not what they claim in conference talks. If you want to know whether Claude is eating into OpenAI's developer share, or whether multi-model routing is mainstream or a power-user quirk, this dataset will tell you. Worth pulling before everyone else forms opinions from it.
|
|
Source →
|
|
---
|
|
03
|
Claude Code is now doing Pokémon team analysis
|
|
|
Thariq posted a thread about using Claude Code to build a personal analysis tool for the game Pokémon Champions. The setup pulls live usage stats from Smogon's npm library, writes code to process matchup data, and generates reports for team planning. It's a one-person research operation for a competitive game, built without a traditional developer workflow.
|
Why it matters: This is a useful illustration of what "AI-assisted coding" actually looks like in practice for a non-professional developer. Thariq isn't shipping a startup. He's using Claude Code the way someone might use Excel with formulas: to answer specific questions faster. The gap between "AI coding tool" and "personal data analyst that happens to write code" is closing fast.
|
|
Source →
|
|
---
|
|
04
|
Microsoft is using AI agents to write formal cryptographic proofs
|
|
|
Microsoft Research published a detailed post on their work verifying Rust code in SymCrypt, the cryptographic library that runs under Windows and Azure. The approach uses Rust for implementation, a tool called Aeneas for verification, and Lean for machine-checked proofs. The new piece: AI agents are writing portions of the proofs themselves, with the proofs then independently verified. Initial release covers SHA-3 and ML-KEM, a post-quantum cryptography standard.
|
Why it matters: Formal verification of cryptographic code is one of those things the industry has always said was too expensive to do at scale. If AI agents can write proofs that a separate system then checks, the cost equation changes. The code securing your bank's cloud infrastructure could plausibly get this treatment within a few years. That's worth paying attention to even if you never read a line of Rust.
|
|
Source →
|
|
---
|
|
|
|
**Swyx is hosting a personal AI demo night in SF tonight** — If you're in San Francisco and building personal agents, Swyx is running a demo event at New Media Lab this Thursday. He notes the last time this meetup ran, a featured speaker got acquired by Amazon's hardware division.
|
|
Source →
|
|
Follow builders, not influencers. A daily digest of what matters in AI.
Read online ·
Archive
|