Cursor

The Cursor forum change removes a way for paying users to inspect consumption and cost information. In a product sold around model access, that makes the bill harder to audit.

HN readers connected the decision to a broader shift: users can move between AI tools more easily than they could move away from a traditional IDE, so transparency can be part of retention.

On Cursor's numbers, Composer 2.5 hits 63.2% accuracy at $0.55 a task — the cheapest in the field, and framed as near-frontier. The presentation even flipped the cost axis so cheapest sits on the right, which several readers found conveniently confusing.

The rebuttal came with citations. On the DeepSWE benchmark — 'probably the hardest to game' — GPT-5.5 gets 64, Opus 4.8 gets 56, and Composer 2.5 gets 16. A developer who's used both models heavily called the parity claim 'absolutely farcical': faster, yes; anywhere near as good, no.

Ravid's claim: a keyboard or a car achieves “tool magic,” becoming an extension of your body, while LLMs are never consistent or fast enough to do that — so you pay conversational effort and get “more code, more tests, more excuses” back.

The counter, from the people the argument forgot: for the deeply introverted, an LLM is the first interface with no social tax at all — no bruised egos, no negotiation, no one to tire out.

The tool is a drop-in proxy across Anthropic, OpenAI, Gemini and open models, scoring each request and routing accordingly, keys encrypted, fully self-hostable. The claim is 40–70% lower cost “with just an endpoint change.”

The objection is structural: agentic sessions live on prompt caching with a five-minute TTL, and the more distinct models you touch, the more cache misses you eat — so a one-planner, one-executor setup already captures the win without the churn.

OpenKnowledge is a local-first markdown editor that auto-syncs a vault over Git and hands editing duties to AI agents. The signature README gif shows an agent drafting a launch recap while you watch.

Then the thread asked the obvious question: it's labeled open source and 'AI-first', so why can it only call Claude, Codex, and Cursor, and not a local Gemma or any OpenAI-compatible endpoint? One veteran wanted 'consequences' for shipping that combination under the open-source banner.

SpaceX's S-1 projects $320bn of AI-services revenue by 2030 at a 74% margin, dwarfing launch and Starlink, which is how the IPO valuation was supported. Its 2025 AI segment was mostly Twitter, with Grok near $0.5bn.

The thread's flip: Hoffman invests in both OpenAI and Anthropic and sits on Microsoft's board, so commenters weigh the vested interest against numbers that still don't add up.

Cursor's founders went from zero to multigenerational wealth in four years, on a product the thread keeps calling a wrapper: an editor fork, a harness, and prompts. Its in-house model, Composer 2.5, is built on Kimi K2.5.

The structure is the catch: this is not a cash deal. SpaceX pays in its own stock, diluting shareholders, while telling IPO investors it sees a $26T addressable market.

And the loudest skeptics can't explain why investors keep funding it anyway.