Earlier this year, I had a great idea.
In my job, I have to memorize a lot of AI companies, so I decided to draw a competitive map on my desk. I divide a piece of paper into columns and list the main ones Companies are determined based on which part of the AI supply chain they belong to.
We had labs like OpenAI, Anthropic, and Google DeepMind that were creating models and chatbots. There are AI coding platforms (Cursor, Cognition, Replit, etc.) and we have built coding assistants that leverage these models. Additionally, startups have emerged building very specialized applications, such as agents that reside within email and help execute marketing strategies or automate tedious tasks such as payroll.
With AI companies relentlessly encroaching on each other’s turf, we’ve noticed that maps are rapidly becoming cluttered.
Continuing rising valuations mean companies need to find new sources of revenue. Becoming a full-stack AI company is especially important to your bottom line as models become rapidly commoditized and big-ticket IPOs loom. Here’s how the terrain is evolving. My paper map is too fast.
Other vibe coders
Companies that start with narrow AI capabilities, such as frontier models, AI agents, and vibecoding tools, quickly expand into one or more other areas.
Last year, Anthropic and OpenAI launched Claude Code and Codex, an AI coding platform comparable to Cursor and Cognition. Now, per-user screenshots of X — Anthropic hasn’t confirmed them — the company may be working on It’s an app builder for non-technical people, and it competes head-to-head with vibe coding stars Lovable, Replit, and Emergent.
For example, Emergent, backed by SoftBank and Lightspeed, says it sees Anthropic coming.
“This is not surprising. We have been expecting this for some time and have been thinking about it internally and preparing for it,” CEO Mukund Jha told me in an interview in April.
Anthropic’s entry into the vibecoding market isn’t game over for the one-year-old startup, he said.
“Coding represents 20% to 30% of the work, and the real pain is getting the application to the last mile,” says Jha.
He added that building secure, production-grade apps is a difficult problem, especially for non-technical users, and may not be solvable in companies that may be “distributed.”
claw rush
In the AI turf war, OpenAI is acting like an agent.
In February, the institute announced it was hiring Peter Steinberger, creator of OpenClaw, an AI assistant builder that surged in popularity earlier this year.
The move adds OpenAI to a growing list of companies entering the agent space, including former Metatech chief Brett Taylor’s Sierra and Salesforce’s AI agent platform AgentForce. Today, Codex has evolved from a coding assistant to a virtual AI agent that can parse and reply to emails, manage files, and schedule meetings.
Smaller players are also excited about this space. Last month, Emergent, which started as a vibecoding platform, expanded into the private agent space.
Other examples of growing overlap include Anthropic moving into the design market and graphic design giant Canva moving into a broader generative AI and productivity suite business.
“Google wanted to touch everything”
If you think you’ve heard this story before, you probably have. However, instead of OpenAI, Anthropic, and Lovable, the characters were Google, Amazon, and Microsoft.
Michelle Cotting, a partner at European venture firm Northzone, said there was a time when FAANG companies had their hands in every pie.
Cotting, co-founder of e-commerce platform Shopping.com, said Google was once a source of great anxiety.
“I remember 25 years ago, when I was building my first company, Google was trying to touch everything,” he told Business Insider in April. “For us at Shopping.com, we had Google launch Froogle, and that’s exactly what we were doing. And we’re like, ‘Oh, we’re done.’
He added, “But as it turns out, it was a side project. How hard are they going to go after it when they’re making so much money in their day job? Well, the answer is they didn’t.”
Tom Sheridan, vice president of early Labable investor RTP Global, agreed that so-called “super apps” — one app to rule them all — are unlikely.
“The talk about super apps is mostly noise and will be resolved by the IPO calendar. Right now we are seeing fundamental model players in the throes of a P&L chicken game,” he said. “Once these companies go public, the cash burn is no longer free and it no longer makes sense to ship to a category that they are good at but not good at.”
‘Google Graveyard’, an unofficial online list, tracks 305 projects that have been decommissioned by the search giant over the years.
Apple is also famous for “sherlocking” new features that make third-party tools irrelevant, but it doesn’t always stick around. In 2023, Apple launched Pay Later, a rival to Klarna and Affirm. It was abolished in 2024.
Cotting said the same thing could happen with OpenAI and Anthropic. These companies were feared by their founders, who feared the day would come when startups would ship applications. I’ve been working on it for months.
It may be more valuable for Anthropic to continue building better models that allow it to charge more for its core services, he said. But if Chinese players and other research institutes become equally good and the model Anthropic may focus more on these services.
In addition to being “Sherlocked,” startups also face another big risk: addiction.
Startups are building multibillion-dollar businesses around APIs that are managed by companies that may ultimately compete with them. For example, Cursor relies on the Anthropic model to power its functionality, but the two also compete as coding assistant providers.
Short-term benefits for customers
More players contribute more Sheridan said the giveaways are a benefit for independent builders and small businesses, but only in the short term.
“Foundation model companies can ship a passable version of almost anything, but if the bundled tools aren’t as good as the specialized tools I’m already using, I’ll revert in one session,” the VC said. “Product sprawl in search of retention bumps risks poor UX, and users know it.”
The sprawl of large labs like OpenAI and Google in all directions also means companies like Reddit and LinkedIn, which host large amounts of data, keep scrapers out. This is bad news for smaller startups like sales technology tools and meeting summarizers, where the data on which to build services is not publicly available.
These changes present opportunities for founders who know what users want from their data.
“With today’s basic model, you can see the recording of a shared meeting for a summary, but you don’t know what folder to file it in, what the team actually cares about, or what follow-up actions are needed,” Sheridan said. “That’s the gap that startups can build on.”
The market is also ripe for consolidation.
“I expect one of the major consumer AI companies to be acquired within the next 24 months, most likely by Google,” he said. “Google has a consumer advertising business to absorb costs, and structurally they are most in need of consumer AI talent.”
Sheridan said the first company to be acquired will get the best price.
“You don’t want consumer AI to be the last one standing when each big buyer probably only gets one chance,” he said.