4作者: throwawayround大约 1 个月前原帖
Posting this because it took me way too long to figure out what was going on, and I wish I had seen a post like this earlier.<p>I just canceled two Cursor Ultra plans. My usage went from a steady ~$60–100&#x2F;month to $500+ in a few days, projecting ~$1,600&#x2F;month. Support told me this was “expected.”<p>I did not suddenly start doing 10x more work.<p>Cursor shows a 200k context window and says content is summarized to stay within limits. Pricing is shown as $ per million tokens. Based on that, I monitored my call count and thought I was being careful.<p>What I did not realise: - Cursor builds a very large hidden prompt state: conversation history, tool traces, agent state, extended reasoning, codebase context. - That state is prompt-cached. - On every call, the entire cached prefix is replayed. - Anthropic bills cache read tokens for every replay. - Cache reads are billed even if that content is later summarised or truncated before inference.<p>So the UI says “max 200k context”, but billing says otherwise<p>Concrete example from my usage:<p>MAX mode: off Actual user input: ~4k tokens Cache read tokens: ~21 million Total tokens billed: ~22 million Cost for one call: about $12<p>Claude never attended to 21M tokens. I still paid for them.<p>This was not just Opus. It happened with Sonnet too.<p>Support explained that this is exactly how the API is billed so there wasn&#x27;t an error and I should just use these models more carefully as they could consume a lot of tokens when they are thinking. But there is a limit to that and what I was charged was way high. There is ZERO transparency about how the cache is used. And the cache breakpoints are decided by Cursor so I don&#x27;t think it&#x27;s fair to throw the ball to Anthropic here.<p>The dangerous part is that cost becomes decoupled from anything you can see or reason about as a user. You think you are operating inside a 200k window, but you are paying for a much larger hidden history being replayed over and over.<p>I am not claiming a bug in Anthropic’s API. This is a product transparency issue. If a tool can silently turn a few hundred dollars of usage into four figures because of hidden caching behaviour, users need much better visibility and controls. Support suggested spend controls but I am actually complaining about how my pre-paid package was consumed.<p>If you use Cursor with long-running chats, agents, or large codebases, check your cache read tokens carefully. The UI will not warn you. The only thing you will see is a few days into your subscription &quot;Your are projected to run out of your usage allowance in a few days&quot;<p>I canceled and moved on, giving Claude Code a shot until this is fixed. Posting so others do not find out the hard way.
4作者: RichHickson大约 1 个月前原帖
I built a macOS menu bar app to track Claude usage in real time via API after hitting limits mid-flow too often.<p>Signed and notarised by Apple. Open source.<p><a href="https:&#x2F;&#x2F;github.com&#x2F;richhickson&#x2F;claudecodeusage" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;richhickson&#x2F;claudecodeusage</a><p><a href="https:&#x2F;&#x2F;x.com&#x2F;richhickson" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;richhickson</a>
2作者: vanpelt大约 1 个月前原帖
Hi HN — I built Catnip, an open-source iOS app that lets you run Claude Code against a real development environment from your phone.<p>Under the hood it spins up a GitHub Codespace, installs Claude Code, and connects the iOS client to it securely. You can use a full terminal when needed, or a lightweight native UI for monitoring and interaction.<p>I built this because Claude Code is most useful when it has access to a persistent environment with plugins, tools, and real repos — and I wanted that flexibility away from my laptop.<p>GitHub gives personal users 120 free Codespaces hours&#x2F;month, and Catnip automatically shuts down inactive instances.<p>Open source: <a href="https:&#x2F;&#x2F;github.com&#x2F;wandb&#x2F;catnip" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;wandb&#x2F;catnip</a> App Store: <a href="https:&#x2F;&#x2F;apps.apple.com&#x2F;us&#x2F;app&#x2F;w-b-catnip&#x2F;id6755161660">https:&#x2F;&#x2F;apps.apple.com&#x2F;us&#x2F;app&#x2F;w-b-catnip&#x2F;id6755161660</a><p>Happy to answer questions or hear feedback.
1作者: xFixItNow大约 1 个月前原帖
I built a “serverless” online chess app and published it on my personal website. It runs mostly client-side and connects players directly over WebRTC (P2P). No accounts, no matchmaking, no backend game server.<p>The entire UI + chess rules&#x2F;validation run in the browser (static HTML&#x2F;CSS&#x2F;JS). For multiplayer, the host generates a 6-digit room code and shares it. The guest pastes it in and connects. WebRTC signaling is done through a small set of endpoints (create room, store offer&#x2F;answer, delete room). After signaling, the connection is direct peer-to-peer via SimplePeer.<p>There&#x27;s a known limitation where connection doesn&#x27;t with both players connected to the same network.<p>Demo: <a href="https:&#x2F;&#x2F;www.adriclumma.com&#x2F;projects&#x2F;chessOnline&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.adriclumma.com&#x2F;projects&#x2F;chessOnline&#x2F;</a><p>Code: <a href="https:&#x2F;&#x2F;github.com&#x2F;ALumma&#x2F;chessOnline" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;ALumma&#x2F;chessOnline</a><p>Let me know what you think!
2作者: backtogeek大约 1 个月前原帖
This idea has been floating in my head for about 10 years. Some of you might remember LowEndSpirit.com back before it became a forum, I started that. I&#x27;ve been obsessed with making tiny, cheap VPS actually useful ever since.<p>TierHive is my attempt to make 128MB VPS great again :)<p>It&#x27;s a NAT VPS (KVM) platform with true hourly billing. Spin up a server, use it for 3 hours, delete it, pay for 3 hours. No monthly commitments, no minimums beyond a $5 top-up.<p>The tradeoff is NAT (no dedicated IPv4), but I&#x27;ve tried to make that less painful:<p>- Every account gets a &#x2F;24 private subnet with full DHCP management. - Every server gets auto ssh port forwarding and a few TCP&#x2F;UDP ports - Built-in HAProxy with Let&#x27;s Encrypt SSL, load balancing, and auto-failover - WireGuard mesh between locations (Canada, Germany, UK currently) - PXE&#x2F;iPXE boot support for custom installs - Email relay with DKIM&#x2F;SPF - Recipe system for one-click deploys<p>Still in alpha. Small team, rough edges, but I&#x27;ve been running my own stuff on it for months. Would love feedback — especially on whether the NAT tradeoff kills it for your use cases, or what&#x27;s missing. (IPv6 is coming) <a href="https:&#x2F;&#x2F;tierhive.com" rel="nofollow">https:&#x2F;&#x2F;tierhive.com</a>