• Intent
  • Posts
  • AI Land Grabs in Code, Language, and Social | Intent 0024

AI Land Grabs in Code, Language, and Social | Intent 0024

Big-tech power plays are rearranging the tools you’ll touch every day – and the incentives hiding underneath are anything but obvious.

Intent exists to help tech talent become more informed, more fluent, and more aware of the forces shaping their careers. We welcome feedback – just hit reply.

The agenda ahead

• OpenAI × Windsurf → why build ($3B!) instead of buy?
• Duolingo’s 148 AI courses → backlash & the fight ahead for AI
• Meta AI breaks out into its own social app — sound familiar?
• 3 interesting startups that just raised money (& might be hiring)

PS. Last call for How to Talk to AI: Work & Productivity. It’s our live, cohort-based, 3-day intensive where we go from how LLMs actually work (Day 1) to a deep-dive on multi-turn, multi-model, multi-modal, agent-powered workflows (Day 2) and wrap things up with frontier models, role-specific magic workflows, and the future of AI (Day 3).

Sherveen teaches the course directly. A note from him: “My goal: give you a mental model around AI/LLMs that lasts 5 to 10 years, not a prompt cheat-sheet that expires with the next OpenAI model update.”

This is the final run at $250 before the price doubles to $500, and Maven’s money-back guarantee means zero downside if it’s not the best AI training you’ve taken. Seats are limited – snag yours now and stay ahead in this new era of work & life. Enroll here!

OpenAI hedges its Copilot dependency

Cursor, the snappy AI IDE (think: an agentic code editor for programmers) beloved by early adopters, just locked down a $900M round at a $9B valuation.

Forty-eight hours later, it’s been confirmed that OpenAI is spending $3B on Cursor-competitor Windsurf (the commercial off-shoot of Codeium) to shore up its own position inside the coding stack.

What’s actually at stake
Most people see autocomplete snippets for code. OpenAI sees a feedback telescope.

Every time Windsurf guesses the next token, it records whether a developer accepts, edits, or rejects the suggestion. Multiply that by millions of daily events and you get the closest thing to a ground-truth dataset for training and evaluating new code-gen checkpoints.

Beyond that, Windsurf gives OpenAI three levers:

  • Distribution – Windsurf can be used standalone but also maintains plugins for a zoo of code editors – JetBrains, VSCode, Neovim, etc – meaning lots of integrated users in the enterprise. Cursor has been focused on its standalone IDE.

  • Dependency hedge – Microsoft continues to be a significant financial backer for OpenAI, and uses its models in key products like GitHub Copilot. If there’s ever a change that leads to Satya cranking prices or nudging usage to other models, OpenAI gets to stay competitive in the agentic coding marketplace.

  • The agentic IDE space is cash-money – Cursor makes $200M in ARR and Windsurf makes $100M. OpenAI won’t mind that the company will immediately contribute to topline revenue given the hungry costs of continued LLM innovation. Building on its own would’ve involved some ramp time.

Our take
It’s all just too early. This is fun to watch, and these tools are fun to use (we’re partial to Windsurf, though we prefer o1 pro in ChatGPT for solving harder coding problems and Claude Code for most feature development).

But — none are a definitive or long-term winner, and so much is subject to change. For now, enjoy watching the big dollar headlines.

Duolingo’s 148 new courses — scaling education or shrinking trust?

Last year, Duolingo confirmed a ~10% reduction in its 2000-strong contractor pool – the same educators who historically wrote and vetted new course content.

Last week, the company unveiled 148 brand-new courses produced largely by a proprietary generative AI pipeline.

But the announcement carried a jarring footnote in the form of an email sent to staff around the same time: hundreds of external course contractors may not have their agreements renewed.

The three-way backlash

  • The human connection – So much of language is cultural, and so much of Duolingo’s success is about its very personalized approach (and friendly owl mascot). The platform shift to AI is whiplash to users who prize the human touch.

  • Labor & ethics – Displaced contributors are drawing parallels to Hollywood’s writer strike: capital captures the upside, expertise becomes a cost center.

  • AI and the future of work – The CEO’s email is earnest in its approach, talking clearly and candidly about how AI is changing all jobs — even his. But this future is uncertain nonetheless, and Duolingo is one of the first companies deciding to talk about it out loud.

Our take
Duolingo got caught in the trap of going first. They’re doing what they should do as a company and as a product organization. AI innovation will make products like this significantly better, cheaper, and more impactful to their end-user.

However, we’re so distracted by politics and rent that we haven’t had the chance to sit around the table as a society to discuss all of the implications of AI on our lives and work. We can’t blame Duo the Owl for that, but we can apparently take our frustrations out on it.

Meta AI, the social experiment

Meta just shipped its first Meta AI standalone app on iOS and Android. On the surface, it’s ChatGPT with Llama 4 under the hood. The twist: chat results default to a Discover tab where anyone can scroll – and remix – public prompts.

Why Meta wants your prompts in public

  • Data flywheel – Public prompt reactions mean free engagement data for fine-tuning conversational tone and meme-speak – the exact kind of juice that keeps TikTok sticky.

  • Network effects – If friends can like, comment, or duet your prompt, you’re more likely to share. That’s a distribution loop OpenAI and Anthropic don’t have today.

  • Monetization runway – Once prompt memes go viral, ad inventory and rev-share follow. A feed is always likely to keep eyeballs glued-on longer than a private chat.

And Sherveen mentioned it two weeks ago in What Could an AI-Native Social Network Actually Look Like?. Remixes might be one of the right ideas to bring LLMs into the social space.

3 interesting startups raised $$

  1. Agree: Free e-signature with invoicing and payment processing on top. They’re giving away DocuSign so they can sell you on Bill.com, a clever combination. $7.2M raised today.

  2. Rainmaker: Drone-based cloud-seeding-as-a-service to nudge rainfall. They make it rain in drought-vulnerable regions and help farmers deal with climate change. $25M raised yesterday. Hiring scientists, mechanics, researchers, electrical engineers, and radar specialists in El Segundo, CA. Link.

  3. Recraft: They’re building an image generation AI platform that’s all about enabling pro designers with controlled styles, brand consistency, and team collaboration features you won’t find in competitor products. $30M raised yesterday. Hiring across engineering, UX, ML, and marketing (remote, London, NY). Link.

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

Sent with Intent,
Free Agency