The Limits Problem Is Getting Worse

Let’s be direct: Claude Code’s usage limits have become a real obstacle, and they’re trending in the wrong direction.
The $20 Pro tier is, at this point, closer to a glorified demo than a working developer tool. But even the $100 Max tier — five times the price — hits walls fast. A recent session involving nothing more than routine refactoring ended with a five-hour lockout after just ninety minutes of work. No override. No graceful degradation. Just a hard stop and a prompt to pay more on top of what you’re already paying.
That’s not a minor inconvenience. That’s a broken workflow.
Anthropic has announced doubled rate limits and a compute partnership with SpaceX, which suggests they’re aware of the pressure. But awareness and resolution are different things, and right now, paying users are the ones absorbing the gap.
What This Looks Like in Practice
Planning your coding sessions around a reset timer is a genuinely ridiculous way to use a productivity tool. And yet that’s where things stand. The limits have tightened progressively over the past several months, and the trajectory isn’t encouraging.
The Alternatives Have Caught Up

For a long time, Claude Code had no real terminal-based competition. Agentic IDEs existed, coding assistants existed, but nothing matched the experience of a proper terminal-native agent with Claude’s reasoning behind it. That gap has closed considerably.
OpenAI Codex — The Quiet Overachiever

Codex had a rocky start. Early impressions were underwhelming enough that most people bounced back to Claude Code immediately. That’s no longer a fair characterization.
OpenAI has been shipping steadily — model improvements, new features, and crucially, Codex is available on the free plan. Its limits are nowhere near as punishing as Claude Code’s current state, which means you can actually use it without scheduling your workflow around an arbitrary timer. For developers who’ve been burned by Claude’s lockouts, Codex is increasingly hard to dismiss.
OpenCode — The One to Actually Recommend Right Now

If there’s a single alternative worth pointing people toward, it’s OpenCode.
It’s completely free. It’s fully open-source. And its defining feature is bring-your-own-model — you can connect it to virtually any popular LLM, or run a local model entirely on your own machine. Plan mode, Skills, agents, MCP server support — the feature parity with Claude Code is striking. The polish isn’t quite there yet, but the flexibility more than compensates.
The practical upside is significant: you’re not locked to one provider, you’re not subject to Anthropic’s next pricing decision, and you’re not getting locked out mid-session. When a better coding model drops next week, you just point OpenCode at it and keep moving.
The Closed-Source Consideration

This one doesn’t affect every developer equally, but it’s worth naming clearly.
Claude Code is a closed-source tool running on someone else’s cloud. For proprietary codebases, client work under NDA, security research, or anything you’d simply prefer to keep off external servers — that’s a real constraint. Anthropic’s data handling may be perfectly fine, but “probably fine” isn’t the same as “auditable and verifiable.”
OpenCode solves this cleanly. The code is open, the behavior is inspectable, and paired with a local LLM, your code never leaves your network. For teams or individuals where data residency actually matters, that’s not a minor footnote — it’s the whole decision.
So, Should You Use Claude Code?

The tool itself is excellent — arguably still the best in its category when it’s actually working. Anthropic’s models remain among the most capable for coding tasks, and the Claude Code experience, when uninterrupted, is hard to beat.
But being the best tool in a category and being the right tool to recommend are two different things. Right now, the limits are punishing, the pricing trajectory is frustrating, and the alternatives have matured enough to be genuinely compelling.
If you’re a light user who values raw output quality above everything else, Claude Code still earns its place. If you’re doing serious daily work, hitting walls mid-session, or handling anything sensitive — Codex or OpenCode deserve a real look.
The best AI coding tool is the one that doesn’t stop you from coding.
Comments (0) No comments yet
Want to join this discussion? Login or Register.
No comments yet. Be the first to share your thoughts!