From 9 items, 2 important content pieces were selected
Technology News
- Gemini 4 Argon draws Hacker News discussion on pricing and capabilities ⭐️ 9.0/10
- 11-year-old soundness bug reported in Rust borrow checker ⭐️ 8.0/10
Technology News
Gemini 4 Argon draws Hacker News discussion on pricing and capabilities ⭐️ 9.0/10
A highly engaged Hacker News thread discusses Google DeepMind's Gemini 4 Argon model release, with 569 points and 439 comments focused on API pricing and generation capabilities. Commenters cite an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off, while questioning the lack of clarity about post-introductory pricing and timing. Discussion highlights the model's SVG output, especially pelican drawings, and its prose quality, with one API user calling it clear and distinct. Others criticize Google's fragmented product experience, including signup, AI Studio and Antigravity discovery, and payment failures for a Gemini AI Ultra package. No official source content is included, so these details remain as discussed by commenters.
hackernews · 737max · Oct 13, 16:03 · Discussion
「Background」 Gemini is Google DeepMind's flagship family of multimodal AI models, and Gemini 4 Argon is its next-generation frontier release, aimed at long-horizon complex tasks across software engineering, finance, legal work, and cybersecurity. Google is offering it at introductory API pricing of $2 per million input tokens and $10 per million output tokens, with later pricing reported at $4 and $20, and the model is not yet generally available. It also supports up to 1 million output tokens, a notable capacity for long-form generation.
「Impact」 For developers using the API, Gemini 4 Argon's introductory $2/$10 per-million-token pricing and strong SVG and prose output make it a notable option, though the post-introductory price remains unspecified in the discussion.
「Community discussion」 Commenters praise the model's SVG pelican outputs and prose quality, with one API user saying it writes clearly and distinctly, while others criticize Google's confusing product lineup, signup and payment friction, and the blog post's buried pricing details. A recurring disagreement is whether image-generation demos or prose quality are the better measure of model capability, with some commenters arguing that well-written prose matters more even if it is less likely to go viral.
References
Tags: #Gemini, #Google DeepMind, #AI models, #LLM, #API pricing
11-year-old soundness bug reported in Rust borrow checker ⭐️ 8.0/10
A GitHub issue in the rust-lang/rust repository (#164250) reports an 11-year-old soundness bug in Rust's borrow checker, drawing significant community discussion about compiler correctness and other open soundness issues. According to the analysis and commenters, the defect appears to be a compiler implementation bug rather than a flaw in Rust's lifetime rules: the compiler should have rejected the affected code under the type system's existing lifetime rules, so the fix is expected to make it reject that code rather than change lifetime semantics. Commenters note that Rust has more than 100 open issues labeled unsound, and that the new trait solver may unblock fixes for some of them, though probably not this one. The discussion also raises the question of whether any real-world crates are affected and whether fixing the bug would cause existing code to stop compiling.
hackernews · formerly_proven · Oct 13, 10:27 · Discussion
「Background」 Rust's borrow checker is the compiler component that enforces the language's ownership and lifetime rules, which is how Rust guarantees memory safety without a garbage collector; a soundness bug means the compiler accepts code that violates those guarantees, potentially permitting behavior such as use-after-free. The borrow checker's implementation has evolved over time — RFC 2094 introduced non-lexical lifetimes, basing lifetime checking on the control-flow graph rather than lexical scopes — and long-standing borrow-checker soundness issues have been filed before, such as rust-lang/rust#51271. Community discussion frames the newly reported issue as a bug in the compiler implementation rather than a flaw in Rust's lifetime rules themselves.
「Impact」 Rust users relying on the borrow checker to reject code that violates lifetime rules are affected, since the compiler accepted code it should have rejected; because the defect is in the compiler implementation rather than the lifetime rules themselves, the fix is expected to simply make the compiler reject that code. It remains unverified whether any published crates depend on the previously accepted behavior, so the practical fallout of the fix is unclear.
「Community discussion」 Commenters largely agree the issue is a compiler implementation bug rather than a flaw in Rust's lifetime rules, with GrantMoyer calling it surprising that it went unnoticed for so long but easy to fix because the compiler should already have rejected the code. Others add context and concerns: weinzierl notes more than 100 open unsound issues and says the new trait solver may unblock some fixes (probably not this one), Eridrus questions whether real-world code is affected or would stop compiling after a fix, and mrkeen points to counterexamples.org for related Rust type-system examples.
References
Tags: #Rust, #borrow checker, #compiler soundness, #type systems, #open source