Platform dependence keeps biting builders
A long list of threads says the same thing. Anthropic stopped Claude Code subscriptions from working with OpenClaw, offering a one-time credit equal to a month's price, and citing 'outsized strain on our systems'. Commenters called it a money grab and said paying for tokens while being told you use too many doesn't add up. An AWS billing bug showed hobbyist accounts owing hundreds of millions to tens of billions of dollars. Docker pulls fail in Spain when Cloudflare gets swept up in La Liga football blocking. Older threads in the same batch carry the same theme: 'Stripe has decided to nuke my entire business', Google Cloud suspending production at 1am on a Saturday, Slack closing its IRC and XMPP gateways.
The key bit: the risk is not outages. It is that one vendor's policy, bug, or legal fight can switch off your business with no appeal. The Debian maintainer's lesson fits here too. If you don't own the thing, it can change without your decision.
The Ask HN on replacing Claude or GPT with a local model shows how builders hedge. Qwen 3.6 27B dense is said to be about Haiku 4.5 level. Gemma 4 on an M4 is too slow. Others run DeepSeek V4 Flash on 2x RTX Pro 6000 at 160 tok/s. Local is workable but still behind the cloud on speed and tooling.
So what?
Treat every API, subscription tier, and cloud account as revocable. Keep a second model provider ready, cap spend with real alerts, and avoid building on a consumer subscription that wasn't meant for automation. A tested fallback is cheap compared with a shutoff.
Read these
Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClaw
AWS: Inaccurate Estimated Billing Data – $1.7 billion
Tell HN: Docker pull fails in Spain due to football Cloudflare block
Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?
Stripe has decided to nuke my entire business
Tell HN: Google Cloud suspended our production projects at 1am on Saturday
Tell HN: Slack decides to close down IRC and XMPP gateways