Search posts, papers, and topics
310 posts · news.ycombinator.com
Android 17 QPR1 is the first Android release since 3.x to introduce new APIs without an AOSP update. The APIs are currently limited to Pixel devices and not yet exposed to other OEMs.
Cloudflare reduced the memory footprint of its Pingora Backend Router by re‑examining the consistent‑hashing implementation in the pingora‑ketama library. By increasing the number of virtual hash points per server from the default 1 to the standard 160 (and applying weighted hashing based on disk capacity), they cut the per‑node overhead enough to reclaim >100 TB of RAM across the fleet. The post…
Apple announced a new iPhone Duo simulator and the Xcode 27.1 beta, but the release notes contain no technical details beyond the headline features.
A US Special Operations analyst used an AI chatbot to augment an intelligence report that falsely claimed a Chinese vessel was carrying nuclear‑related cargo. The report triggered a near‑miss military operation that was aborted once the hallucination was discovered. The incident highlights the lack of verification standards and the risk of relying on AI‑generated intel for targeting decisions.
Researchers used photon‑emission microscopy to locate the DEBUGEN register on a decapsulated RP2350, then applied focused laser pulses to flip its bits, re‑enabling Secure debug and extracting a secret stored in OTP memory. The attack needs physical access and a $250 k lab setup, showing that the chip’s permanent debug‑disable can be overridden.
A new study shows that the forebrain and hindbrain develop from distinct progenitor cells (OTX2‑expressing and GBX2‑expressing) in mouse, human, chicken, zebrafish and acorn worm embryos, suggesting the vertebrate brain originated from the merger of two primitive nervous systems over 550 M years ago. This insight enabled the first successful in‑vitro generation of human hindbrain motor neurons.
The author used Claude to generate a Lean formalisation and proof of Conway’s refinement conjecture for omnific (surreal) integers. The proof passes mechanical checks and informal review but lacks independent verification. The post narrates the workflow, token cost, and reflections on AI‑assisted theorem proving.
Cloudflare Quick Tunnels let you expose a local service via an outbound‑only connection to Cloudflare’s edge network. The service gets a public URL (anycast across 335+ cities) with automatic TLS, DDoS protection, and no inbound ports. It’s positioned for short‑lived agents, webhooks, and testing loops, with JSON health output and no config files.
The 2nd U.S. Circuit Court upheld the border-search exception, allowing agents to seize and examine travelers’ phones without a warrant, rejecting amicus arguments that First and Fourth Amendment rights require a warrant.
Passkeys improve phishing resistance but introduce higher lock‑out risk for individuals. Hardware keys have low per‑device account limits and no backup, while synced passkeys tie you to Apple/Google and can be lost if the provider bans you. Third‑party managers are still immature. For most users, a strong password + TOTP combo remains safer than current passkey ecosystems.
The author argues that “vibe‑coding” with LLMs can cause engineers to reinvent existing solutions, using Bend 2—a language that asks LLMs to write long formal proofs—as a case study. By re‑implementing the same demo in SPARK/Ada, the author shows that established formal‑verification tools achieve the same guarantees with a tiny proof, highlighting the need for domain research before prompting LLM…
New unredacted filings in the NYT v. OpenAI/Microsoft case show internal admissions that the companies scraped paywalled news content at massive scale, stripped copyright notices, and acknowledge that their products cannibalize traffic, raising serious fair‑use and copyright concerns.
OpenJev is a browser‑only demo that lets you load small LLM checkpoints (e.g., MiniCPM‑5 2B, Qwen3 0.6B) onto your GPU and compare two inference paths: reading raw logits for a set of options versus prompting the model to emit a JSON with option probabilities token‑by‑token. The page reports model sizes, download times, balanced accuracy on a few benchmarks, and wall‑clock timings measured with `…
ZCode (Zhipu’s AI coding desktop) silently archives your entire workspace—including full .git history, LFS cache, and config—encrypts it with a server‑supplied RSA public key, and uploads the ciphertext directly to Aliyun OSS. The upload runs unconditionally for any logged‑in user, cannot be disabled via UI, and the decryption key lives only on the server, giving Zhipu full read access to your co…
