Claude Keeps Extending Its Free Code Limits. Here's Why That's a Red Flag.

Anthropic just extended Claude Code's promotional usage limits to August 19. Again.

This is the third or fourth extension now. I've lost count. Every time the deadline approaches, they push it back another month.

At first, I thought this was a nice promotional move. A way to get more developers hooked on Claude Code before the limits kicked in properly.

Now I'm reading it differently.

This isn't a promotion. It's a retention campaign.

Every time Anthropic extends the free limits, they're admitting something uncomfortable: people leave when the free stuff stops.

The goalposts keep moving because the moment they don't, users start looking elsewhere. And right now, "elsewhere" is getting increasingly attractive.

The competition is getting real

Two things have happened in the last few months that I think are directly responsible for these extensions.

First: OpenAI keeps giving away free usage.

ChatGPT's free tier keeps expanding. More prompts, more capability, no credit card required. When free is good enough for most tasks, asking people to pay $20/month for Claude Code is a harder sell. The habit is being formed on OpenAI's turf.

Second: Kimi K3 happened.

This is the one that seems to have genuinely spooked Anthropic. Kimi K3 is an open-source model that, on some tasks — including frontend design — is apparently outperforming even Claude Fable. Think about what that means: an open-source model, cheaper to run, no vendor lock-in, matching or beating Anthropic's flagship creative model.

When your differentiator is also someone's commodity, you have a problem.

What this looks like from the inside

I talk to a lot of developers. The pattern I'm hearing:

1. They tried Claude Code when it was free/cheap
2. They liked it
3. When limits kicked in, they didn't upgrade — they switched back to OpenAI for everyday tasks and kept Claude Code for "serious" work only
4. The serious work isn't enough to justify the subscription on its own

That last point is the key one. If Claude Code isn't people's daily driver — if it's just the "specialist" tool they open once a week — the $20/month subscription is hard to defend to yourself, let alone to a finance team.

"Claude is in a very hard situation. They have to keep giving free usage so they don't lose people."

That's a direct quote from someone who's been watching this space closely. I think it's right.

What happens when the extensions stop?

That's the experiment Anthropic is running. Every extension is a test: what happens if we let the promo expire this time?

So far, the answer has been "churn." Not catastrophic, probably. But enough to make them push the date back. Again.

The real test will be when they finally let it expire and the limits actually kick in at full strength. My guess: we'll see a meaningful drop in active Claude Code users, and Anthropic will have to decide whether to cut their losses or compete on price in a way they clearly don't want to.

The broader picture

This is what the AI subscription wars look like in practice. Not the polished PR announcements — the messy reality of usage limits being extended, free tiers being matched, and every company trying to buy user habit before the competition does.

OpenAI has the retail user base. They already won the "make it a habit" battle for most people.

Anthropic is fighting that war on two fronts: against OpenAI at the low end, and against open-source models like Kimi K3 at the cost frontier.

The extensions won't stop until one of those battles resolves. I don't think either is resolving soon.

---

If you want more breakdowns of what the major AI players are actually doing — not just announcing — I post those on LinkedIn. Follow me there for the play-by-play.