A Week of Confrontation: Trump & Musk (Obviously) + OpenAI & Anthropic Collide on Coding
Plus, highlights from the famed Mary Meeker's latest "Trends" report
This Week in Short
OpenAI’s move to buy Windsurf prompted Anthropic to cut off the coding app — a sign of whats to come. One thing we can be confident of is crazy AI growth, per Mary Meeker’s “Trends” report. A wave of IPOs continued with a big pop for stablecoin provider Circle on Thursday, and M&A is picking up too. AI video is suddenly everywhere as both Hollywood dealmaker Jeffrey Katzenberg and Meta back tools for AI video ads.
The Main Item
Coding Apps Are a Test-Case for LLM Strategies
An existential question hangs over developers of AI applications: will the big LLM companies smother them in their cribs?
Welcome to the age of AI power plays.
The current machinations around coding apps offer a preview of how this may play out across the industry. OpenAI kicked things off with its recent move to buy Windsurf, a leading coding app that makes heavy use of Anthropic LLMs. Anthropic this week responded by cutting off Windsurf’s access to its more advanced models — hardly a surprising development, given the fierce rivalry between Anthropic and Open AI.
“I think it would be odd for us to be selling Claude to OpenAI,” Anthropic co-founder Jared Kaplan said at a TechCrunch event Thursday.
Brought to you by Forethought
CX used to be built in pieces. But today, companies like Scale, Cohere, and Airtable use Forethought to connect every part of the customer journey.
Now, a team of AI agents can work across sales, onboarding, support, and retention. They understand, plan, take action, and improve.
No circles. No silos. No dead ends. This is what enterprise AI should look like.
Windsurf Denied Access to Claude
Anthropic’s actions prompted some performative indignation from Windsurf, which contended they would “harm many in the industry.” But the situation highlights the quandary facing many AI app-makers, one that’s similar to what developers faced with Windows back in the day, and more recently with Facebook and Google and Apple: If you’re building on top of someone else’s platform, you’re vulnerable to being “Sherlocked,” as they call it in the Apple world, i.e. seeing the platform provider do whatever it is you’re doing themselves, or handicap you for some other reason.
Investor and lobbyist Zak Kukoff pointed out another example in OpenAI’s announcement of a “record” mode for ChatGPT on Mac versions of Teams, with further roll-outs to come. “Granola getting Sherlocked by OpenAI,” he wrote. “At some point model providers are going to need to decide if they want to be stable platforms or compete for every vertical. Platform risk has never been higher.”
The comment thread reflects the industry tensions over this issue, and disagreements over how it will play out.
OpenAI, Anthropic, Google-Gemini, and xAI have easy mechanisms for putting the squeeze on developers. They can limit access, of course, and they can also apply pressure through pricing, where there are few guidelines and little experience on what things should cost. Like the big platforms of previous eras, they seem very capable of taking over numerous verticals, and when it comes to major ones like coding, they would appear to have every incentive to do so.
That might be especially true of Anthropic, which lacks a viral consumer product on a par with ChatGPT and has leaned into business applications for revenue. On the flip side, Anthropic’s emphasis on recruiting developers to use its APIs makes it even more imperative for the company to maintain good relationships.
This is the key constraint on LLM providers going big with their own applications: they can’t serve every vertical or specific use case, and they need a robust developer community to be building on their platforms. Veteran VC and analyst Mary Meeker, in an interview about her Trends forecast (highlights below), pointed to a 1999 video in which then Microsoft CEO Steve Ballmer, pacing the stage with his shirt soaked in sweat, bellows “developers, developers, developers” over and over again, in answer to what would be the key to the company’s success on the Internet.
“He's right, that the companies that get the best developers often win,” Meeker said.
The intensifying power struggles in the AI industry aren’t limited to platforms versus developers either. Even as it was pushing back on Windsurf, Anthropic was being sued by Reddit, which accused the company of vacuuming up its content for model training without permission. The industry’s battles over IP, while ultimately a legal issue, also reflect the power dynamics as companies weigh the risks and benefits of licensing versus litigation.
In the current environment, the big LLM vendors are mostly in the driver’s seat, but self-interest dictates that they move thoughtfully. They have to compete with one another for developers, for one thing, and open source models are also a threat.
From where we sit, it seems likely that the LLM providers will in fact ultimately own at least some of the major applications categories; they have an enormous need for revenue to cover their infrastructure costs, and may not be in a position to share the profits with third party apps on everything.
And yet, even as Anthropic and OpenAI wrestle with Windsurf, the other big coding app, Cursor, was being touted as the fastest-growing startup ever on Thursday as it raised a fresh $900 million at a valuation of almost $10 billion. Whether it ultimately gets bought or stays independent, that’s already a huge win.
For the LLM companies, finding the right balance in their relationships with developers, as well as other players across the ecosystem, will be one of their biggest tests. Developers, for their part, will need to be shrewd in anticipating when to race ahead — and when to get out of the way.
Cerebral Valley Podcast
AI Agents Are Already Here
Eric joins Volley’s Max Child and James Wilsterman for our first episode of the Cerebral Valley Podcast ahead of our June 25 summit in London.
Our predictions from last year around the scaling wall look totally naive just a few months later. ChatGPT’s Deep Research and AI-powered coding assistants are bringing our agentic future even closer to reality — check out Thomas Ptacek’s in-depth exalting of AI progress around software engineering for further proof. Other predictions around AI filmmaking are looking a bit more close.
One Big Chart
Startup M&A Activity Picked Up After Wiz Deal
May was the best month for startup M&A in 10 quarters, if you leave aside the March numbers that were skewed by Google’s $32 billion acquisition of Wiz. May saw over $24.7 billion in deals for venture-backed startups, including OpenAI’s acquisitions of Windsurf and Jony Ive’s hardware startup Io.
Other May deals that moved the numbers include Databricks’ purchase of Neon, Coinbase’s move to buy Deribit, and DoorDash’s acquisition of SevenRooms and Deliveroo.
AI Trends
5 Slides from Mary Meeker’s Big AI Report
The veteran Bond Capital VC who rose to fame during the first internet boom for her prescient forecasts of the meta-trends has dropped her latest “Trends” report, a weighty 340 slides in all. The TL;DR is easy enough: generative AI is even more earth-shaking than you think, and the speed of adoption has made the internet look like a tortoise by comparison. We’ve pulled a few slides we found especially interesting below. We were struck by the way in which the Internet has enabled AI to land everywhere all at once around the globe — a big change from prior tech waves.
