Journalists start vibe coding and why Apple picked Google over OpenAI | Intent, 0029

It’s been a few months since we last landed in your inbox, but we’re still alive!

As a reminder, Intent is all about helping talent in tech become more intentional with their career by staying informed, fluent, and aware of what’s going on in and around the industry. Thanks for sticking with us!

Today's agenda:

• Journalists are getting Claude-Code-pilled

  • It’s a very loud signal about AI-enabled “personal software”

• Apple picked Google over OpenAI for Siri: why?

• Grok’s deepfake mess is turning into a real-world AI regulatory test

If you weren’t paying attention to Claude Code already…

Here’s a signal that the times are a-changin’: journalists and tech outsiders are starting to get into agentic coding tools. And they aren’t just trying tools like Claude Code once or twice, but writing about them as new obsessions.

• Bloomberg’s Joe Weisenthal has been posting about his new software side project, Havelock, which detects the presence of orality in text. And the experience of building and iterating on it with Claude Code over the past few weeks has him writing newsletters like “AI’s Productivity Potential Has Never Been More Obvious.” (Bloomberg, paywall)

• Mega-Substacker Matt Yglesias has been using Claude Code to build charts and aggregate data, blaming it (in jest) for a dip in his writing productivity.

• Casey Newton (of Platformer and NYT’s Hard Fork) recently wrote about how building his personal website turned him into a Claude Code believer.

• And Anthropic released a gimped version of Claude Code the other day called Claude Cowork, meant to more specifically appeal to non-tech workers.

There are a variety of lessons that are intermingled here, including:

• The beginning of an era of personalized software, when even non-technical folks will be building micro-apps for their personal and professional lives.

• The continued commoditization of SaaS, where individuals and SMBs will start to build their own custom solution to replace their pricey software subscriptions.

• The coding agents are reaching the mainstream: both in how often they’re discussed, and how easy they are to try.

So, if you aren’t yourself already hopped up on this trend… you should be using Claude Code (or its counterparts — we prefer Codex CLI lately) ASAP!

And if you’re a builder, know that far more Weisenthals and Yglesiases are out there looking for more curated solutions to their use case, who will be just as amazed when you help them realize the agentic AI capability that’s been unlocked over the past 12 months.

PS — Build Personal Software: Claude Code for Non-technical Users, free 30min workshop on January 26, sign up here!

Apple picks Google to power Siri

The other day, Apple and Google announced a long-term agreement to have the Gemini class of models power Siri and other AI modalities within Apple’s product suite.

Apple’s existing integration with OpenAI (rolled out in late 2024) is still there, but expect it to go from an opt-in “complex queries” layer to being deprecated as this Google partnership takes hold.

Why would Apple choose Gemini over OpenAI?

We don’t know for sure, but we have a pretty good guess beyond corporate vibes and partnership chemistry: the product physics match up well here. And it mostly comes down to latency.

Apple’s phones target “normies” and people who otherwise expect Siri’s rapid responses. Latency is super important for these use cases.

While OpenAI has best-in-class models (and, in our opinion, the smartest class of products), they aren’t fast — and OpenAI’s fast models aren’t very smart. Google’s Gemini 3 Pro, on the other hand, is pretty darn fast while being #2 or #3 in intelligence. And Gemini 3 Flash is really darn fast while being decently smart.

Plus, Gemini models are more expansively multimodal (phone, camera, voice operability), which fits the phone OS integration that Apple will inevitably desire.

So, did Apple bet on Google because it thinks Gemini is or will be better? Maybe, but also maybe not — given there’s reporting (FT, paywalled) that OpenAI may have been the one to have soured on the deal, it might be the product physics above that made Sam Altman think this might not be a match made in AI heaven.

Elon Musk continues to be the world’s largest hypocrite

You’ve probably already heard that Elon Musk’s Grok has been at the center of a genuinely nasty trend: people using its image generation capabilities to generate non-consensual sexualized edits and deepfakes. Inevitably, women and minors were heavily targeted by a waterfall of deepfake harassment on X.

xAI has taken steps to limit this functionality over the past few days, but mostly by limiting it to paying users and restricting what the bot can do publicly (though, Grok remains more free-wheeling in the private version available via its dedicated app).

Governments from across the world launched bans and investigations, and we’re seeing regulators and press outlets start to pay serious attention.

Beyond the grotesque outcomes here that Elon and his team are enabling, the hypocrisy is almost more stunning: Elon has come out and repeatedly defended the AI’s behavior (The Verge, paywall), blaming the infringing users, referencing free speech, and even amplifying the images and videos that he himself finds personally arousing and acceptable.

If we rewind to 2024’s controversy around Gemini’s first significant image model, Elon was one of many on the tech right (Axios, paywall) yelling that it was unacceptable for Google’s models to be so woke so as to only generate images of black founding fathers.

It was likely a system prompt that over-corrected for diversity, but he and the tech right accused Google of being “ideologically broken.” Yet, when Grok generates non-consensual deepfakes, it’s everyone else’s fault.

It’s no wonder his own partners and mothers to his children, Grimes and Ashley St. Clair, are calling him out publicly after being personally victimized by his platform — but it’s a stark reminder that AI safety in 2026 is only going to be weaponized at-will to serve the agenda of specific tech oligarchs.

Not that we should be surprised at this point.

Two quick startups to watch (and are hiring!)

• Oboe’s raised $16M to build an AI-powered learning platform that builds on-demand micro-courses (kind of like a different flavor of NotebookLM) — the team previously built Anchor, a podcasting platform it sold to Spotify, and is hiring across roles in NYC.

• Numeric’s raised $51M to build AI for accounting, one of the clearest ROI use cases for enterprise AI in existence and one of the most “near-term possible” industries that’ll be completely turned over by AI in the mid-term. Hiring for engineering and GTM roles in SF and NYC.

Think a friend could use a dose of Intent? Forward this along – inbox envy is real.

Sent with Intent,
By Free Agency