Dev Tools Just Collapsed Into Three Players. You're Not Ready.

Your Toolchain Belongs to Three Companies Now

Google just unified its AI coding tools under the "Antigravity" brand. IDE, model, deployment pipeline. One throat to choke, one vendor to depend on, one company that also controls the AI writing your code. If you are a CTO running a Swedish engineering org and you did not feel a chill this week, you are not paying attention.

Here is what is actually happening: the independent developer tools market is entering an extinction event. Not slowly. Not hypothetically. Right now, in May 2026, the three companies that control the foundation models, Google, OpenAI, and Anthropic, are each building complete vertical stacks that bundle coding agents, IDEs, deployment, testing, and monitoring. They are giving these away for free or near-free because the tools are not the product. You are the product. Your data, your codebase, your engineering patterns. That is what trains the next model.

And the best-of-breed indie tools your team loves? Cursor just topped Gartner's AI Coding Agents quadrant. Congratulations. That is like being named the best buggy whip manufacturer in 1908. When OpenAI and Google bundle competing features for free inside their own platforms, the standalone AI coding agent has the life expectancy of a fruit fly.

I run a tech company in Jönköping. We build AI agents and SaaS products for clients across Europe. I have to make real decisions about this stuff every week. So let me tell you what I see, what I think it means, and what you should actually do about it.

The Vertical Integration Play Is Obvious. Why Is Everyone Surprised?

The pattern is not new. Microsoft did it with Azure plus GitHub plus VS Code plus Copilot. That was the dress rehearsal. Now Google is doing it more aggressively with Antigravity, connecting Gemini models directly to Cloud, to the IDE, to deployment. OpenAI is building out its own coding agent that lives inside its ecosystem. Anthropic has Claude Code and is pushing hard into agentic developer workflows.

Each of them wants the same thing: to be the platform your entire software development lifecycle runs on. Not a tool in your stack. The stack itself.

This is not about AI agent development benchmarks. SWE-bench scores are a distraction. The real competition is over who owns the surface area of your engineering process. Whoever owns that surface area gets to see every line of code your team writes, every bug pattern, every architectural decision. That data makes their models better. Their better models make their tools stickier. Flywheel. Done.

The indie tools, the Cursors, the Windsurfs, the standalone coding agents, they are in a brutal position. They do not control the model layer. They pay API costs to the same companies that are now competing with them. Their margins get squeezed from both sides. Some will get acquired. Most will just fade.

Sweden Is Particularly Exposed

Swedish engineering culture loves best-of-breed tooling. We have a tradition of picking the sharpest, most elegant tool for each job. A CI system from here, an observability platform from there, a deployment tool from that cool Berlin startup. Our Stockholm and Gothenburg engineering teams run beautifully assembled stacks of 15 to 20 independent tools, each chosen with care.

That was a strength. It is now a vulnerability.

When the big three collapse the toolchain into integrated platforms, the switching cost for a Swedish org running indie tools is enormous. You are not just swapping one tool. You are rearchitecting your entire development workflow. And the longer you wait, the more your competitors who jumped on the platform early benefit from the compounding advantages of integrated AI assistance.

SvD ran a piece this week with the headline "AI-verktyget knäcker kodare: Ont i magen." AI tools are giving developers stomachaches. And Computer Sweden is reporting that experienced Swedish developers are now vibe-coding more than juniors. Think about what that means. Your senior engineers, the ones you depend on for architectural judgment, are already deeply coupled to these AI coding tools. The dependency is not future tense. It is present tense.

Meanwhile, Swedish software development organizations have almost no leverage over these platforms. We are consumers, not stakeholders. When Google changes Antigravity's pricing or data policies, your Malmö engineering team gets the email the same day everyone else does. No input. No negotiation.

The Security Problem Nobody Wants to Talk About

This week also brought a VS Code supply chain attack that hit GitHub, OpenAI, and Mistral AI. Let that sink in. The companies building the AI coding tools got compromised through the very extension ecosystem they are asking you to trust with your codebase.

As these platforms vertically integrate, the attack surface does not shrink. It concentrates. One compromised extension or model update in a platform that controls your IDE, your CI/CD, your deployment, and your monitoring means an attacker potentially has access to everything. Not one tool. Everything.

Swedish companies operating under GDPR and the EU AI Act have regulatory obligations around data handling and AI transparency that these platforms are not designed to help you meet. They are designed to ingest your data, not protect it. The compliance gap is real and it is growing.

I keep seeing the system prompts and internal tools of major AI coding agents get leaked on GitHub. Over 138,000 stars on that repo. The entire industry's secret sauce is sitting in plain sight, and what you see when you read those prompts is that these tools are deeply integrated into your workflow in ways most engineering leaders do not fully understand.

Where This Goes: 2027 to 2030

Let me lay out what I think happens next.

By end of 2026: Google Antigravity, Microsoft/GitHub Copilot Workspace, and Anthropic's agent platform are the three dominant developer environments. Cursor either gets acquired or starts bleeding users. Most standalone AI coding tools lose 40% of their user base within 12 months.

By 2027: These platforms are not just writing code. They are making architectural decisions, choosing frameworks, configuring infrastructure. The "AI agent" is not an assistant anymore. It is a team member that happens to be controlled by the same company selling you cloud compute. MVP development timelines drop from weeks to days, but the lock-in becomes nearly total.

By 2028-2029: As we approach more capable AI systems, the platform that controls the developer toolchain essentially controls software production. If one of these companies achieves something close to AGI-level coding ability, the platform advantage becomes permanent. You will not switch, because switching means retraining the AI on your entire codebase, your patterns, your business logic. The cost is prohibitive.

The regulatory gap: The EU AI Act was written for a world of identifiable AI systems doing specific tasks. It was not written for a world where your entire software supply chain runs through a single vertically integrated AI platform. Brussels is at least two years behind on this. Stockholm is further. The Swedish government still thinks AI policy is about ethics committees and research grants. It is not. It is about market concentration in the infrastructure your entire digital economy runs on.

The Open Source Counterweight

There is one force that could slow this down: open source. Projects like Opencode, the open source coding agent that is trending hard on GitHub right now, represent an alternative path. N8n, the fair-code workflow automation platform with close to 190,000 stars, shows that self-hosted, open alternatives to platform-locked AI tooling are possible and popular.

But I am going to be honest: open source alone will not win this. The big three have the compute, the data, and the model advantage. Open source can be a hedge, a way to keep optionality, a pressure valve against total lock-in. It cannot be your entire strategy. Not if you want your team shipping at the speed your competitors on the integrated platforms are shipping.

What to Look At

Here is what I would put on your radar this week if you are making toolchain decisions:

1. Opencode: An open source coding agent. Not a replacement for Cursor or Copilot today, but the kind of project that matters as a long-term hedge against platform lock-in. Self-hostable. Worth evaluating now.

2. ECC (Agent Harness Performance Optimization): Nearly 189K stars. A system for managing skills, memory, and security across Claude Code, Codex, Cursor, and beyond. If you are going to use multiple AI coding agents (and you should, for now), this kind of meta-layer is how you avoid single-platform dependency.

3. n8n: Self-hosted workflow automation with native AI capabilities. The kind of plumbing that lets you build AI-augmented development workflows without handing everything to one platform. We use it at HEIMLANDR for exactly this reason.

4. Your own audit. Sit down. List every tool in your development pipeline. For each one, ask: which of the big three platforms could replace this for free tomorrow? If the answer is "all of them," that tool is dead. Plan accordingly.

What We Are Doing at HEIMLANDR

I will tell you what we decided here in Jönköping. We are treating the big three platforms as compute and model providers, not as our development environment. We build our own tooling layer on top. We keep our AI solutions architecture portable. We use open source where we can, proprietary where we must, and we never let a single vendor see our entire pipeline.

Is this more work? Yes. Does it cost more in the short term? Yes. But we are a software development company in Sweden. Our independence is our product. The moment we are fully locked into Google's or Microsoft's or Anthropic's platform, we are not a tech company anymore. We are a reseller.

That is not a position I am willing to be in. And if you are building anything serious, it should not be a position you accept either.

The Challenge

Here is what I want you to do this weekend. Open your engineering team's tool inventory. Count the independent tools. Then count how many of those tools have a direct competitor being built into Google Antigravity, GitHub Copilot Workspace, or Anthropic's agent platform. Do the math. Then ask yourself: when those indie tools lose 60% of their revenue because the big platforms bundle the same functionality for free, who maintains the tool your production pipeline depends on?

If the answer makes you uncomfortable, good. Discomfort is the start of good decisions. The toolchain consolidation is happening whether Swedish engineering orgs prepare for it or not. The ones that plan now keep their options. The ones that wait get to learn what platform lock-in really feels like.

Build things. Own your stack. Stay sharp.

Fredrik Brunnberg is the CEO of HEIMLANDR.IO, building AI and software solutions from Jönköping, Sweden. This is the daily HEIMLANDR briefing. If you found this valuable, share it with someone who builds things.