Three months ago, there were six major AI coding assistants. Now there are 20. By next quarter, there will be 30. Most of them will be dead by the end of the year.
Not because the technology doesn't work. Because everyone's building the same thing, racing toward feature parity, mistaking a crowded space for a mature market.
Developer Ravi Patel argues the AI coding space is in explosion mode, not consolidation. The assumption that tools will merge into platforms is too early. Most current tools haven't figured out what they're actually good at - they're just copying each other's roadmaps.
The Feature Parity Trap
Here's what's happening: Tool A launches with inline code completion. Tool B adds inline completion plus chat. Tool C adds completion, chat, and repo-wide search. Tool D does all three plus debugging. Everyone's playing feature leapfrog, assuming the winner is whoever has the longest feature list.
That's how you end up with 15 tools that all do the same eight things with slightly different UX. None of them own a specific use case. None of them are dramatically better at the thing developers actually need most. They're platforms in their own minds, wedge tools in reality.
Patel's thesis: the survivors won't be the ones with the most features. They'll be the ones that own a wedge - a specific dimension where they're measurably, undeniably better. Latency. Cost. Integration depth. Specialisation in a language or framework. Something concrete that matters more than a bullet point.
What Developers Actually Optimise For
Speed matters more than capability. A tool that suggests decent code in 200 milliseconds beats a tool that suggests perfect code in 2 seconds. The fast one stays in flow. The slow one breaks concentration. Developers will choose the fast one and fix the code themselves.
Cost matters for teams. A tool that's 80% as good but half the price wins at scale. Individual developers might pay for the best. Engineering managers buying for 50 people do maths. The cheaper tool that's good enough gets the contract.
Integration depth matters for existing workflows. A tool that reads your team's style guide, understands your internal libraries, and knows your deployment patterns is worth more than a general-purpose model that's technically more powerful. Context is leverage.
The Explosion Phase Looks Like This
Right now, every AI lab, every dev tool startup, and every cloud platform is launching a coding assistant. They're all raising funding on TAM slides that show "every developer in the world". They're all promising to 10x productivity. They all have a chat interface and autocomplete.
This is normal. It's what explosion phases look like. Lots of capital, lots of attempts, very little differentiation. The market hasn't figured out what it wants yet because there are too many options to evaluate properly. So everyone hedges by building a bit of everything.
Patel's point: this isn't the end state. It's the messy middle. Consolidation comes later, after the wedges get defined. After someone owns latency so completely that nobody else can compete there. After another tool becomes the standard for Python developers specifically. After cost-optimised models prove they can serve teams at 1/10th the price of premium tools.
Who's Positioned to Survive
GitHub Copilot has distribution - it's already where developers work. That's a wedge. Cursor has speed and UX polish - it feels faster than competitors, even when it isn't. That's a wedge. Replit has the full environment - you code, deploy, and host in one place. That's a wedge.
The tools trying to be "Copilot but better" without a specific dimension of better? Those are the ones that die. Better at what? Faster? Cheaper? Smarter for which use case? If the answer is "generally better", that's not a wedge. That's a hope.
The local-first models (running on-device, no cloud) have a wedge if privacy-sensitive companies adopt them. The ultra-cheap models have a wedge if they can serve cost-conscious teams. The framework-specific tools (Rails-optimised, React-native, embedded systems) have wedges if they go deep enough.
What Happens Next
Expect more launches, more funding announcements, more feature parity. Then expect the collapse. Tools that couldn't define their wedge will quietly shut down or pivot. Acquihires. Shutdown announcements framed as "mission accomplished".
What survives will be smaller and sharper than what exists now. Not platforms. Not general-purpose assistants. Wedge tools that own one thing completely. And maybe - maybe - one or two actual platforms built by companies with distribution and capital to subsidise the land grab.
The merging take assumes we're at the end. Patel's right - we're still at the beginning. The market hasn't settled. The wedges haven't formed. We're still in the part where everyone thinks they can be everything. That phase always ends the same way.