DeepSeek has open-sourced DeepInfer, the high-performance inference engine behind its production services.
Debate over whether inference margins are sustainable, with camps citing 70%+ profits vs. commodity commoditization and missing training costs.
2273 pts · 873 comments
An upcoming critical PHP security release addresses vulnerabilities (CVE-2026-109618 and -109619) affecting versions 5 through 8.
Commenters describe the vulnerability as catastrophic, possibly worse than Log4Shell, with a PoC already in private hands.
1345 pts · 563 comments
Private keys of eight eMudhra intermediate CA certificates were compromised, allowing website impersonation.
Web PKI's universal issuance model drew heavy criticism; DANE and name constraints were proposed as fixes.
1095 pts · 424 comments
Mistral launches Voxtral 3, a voice AI family with STT, TTS, and an SDK for real-time agent pipelines.
Closed model choice disappoints; local alternatives like Parakeet and Kokoro recommended.
727 pts · 180 comments
iPadOS 27 breaks the ability to present one app while using another on iPad.
Two camps: iPads are not real computers versus iPads work fine in specialized settings.
586 pts · 286 comments
UNEP declares the Paris Agreement's 1.5°C goal no longer achievable.
Critics blame uncompetitive night trains and misguided European energy policy for stalling progress.
533 pts · 600 comments
Browser vendors are collaborating on Interop 2027 to standardize CSS functions, corner-shape, and anchor positioning.
Two camps: those welcoming CSS's growth, those warning against turning it into an awkward programming language.
487 pts · 202 comments
OpenBSD 8.0 adds hibernate support for arm64 and expands kernel virtual address space on amd64.
Split on the satirical AI artwork; praise for the simple sysupgrade process.
479 pts · 155 comments
Facebook will require pages to pay for posting more than two links per month.
The thread largely condemns the policy as a cash grab that hurts small businesses and pushes users off-platform.
452 pts · 394 comments
A soon-to-be father rethinks education's purpose as AGI threatens the value of programming skills.
Corrections to Gates/Einstein myths and a split on whether schools teach genuine skills or just obedience.
442 pts · 541 comments
eMudhra's CA private keys were single-byte numbers like 4, recoverable in milliseconds from public certificates.
Outrage at 8-bit keys, with speculation about script errors and lack of basic validation before trusting CAs.
432 pts · 199 comments
Cloudflare ported TypeScript's Go compiler to Rust, achieving 3-4x speedups and enabling WASM-only sandbox execution.
AI-generated port sparks debate over speed claims, WASM motives, and whether zero-human-code compilers are trustworthy.
424 pts · 206 comments
MicroG provides a free reimplementation of Google's proprietary Play Services for Android.
Sentiment is positive but split on security: some recommend GrapheneOS's sandboxed Google Play over MicroG.
414 pts · 121 comments
Ofcom admits it cannot enforce Online Safety Act fines in the United States.
Commenters view the outcome as a straightforward sovereignty issue and warn of UK censorship escalation.
408 pts · 341 comments
A Google Cloud outage in europe-west2 disrupted major European banks and financial services on Monday.
Single-region cloud dependency blamed for outages; Monzo and Starling praised for multi-cloud resilience.
394 pts · 200 comments
Hacked AfD emails reveal antisemitism, racism, and violent fantasies.
Frustration with the paywall dominates, as commenters share alternative sources.
382 pts · 543 comments
333 Tor exit nodes were compromised in an SSH MITM attack using the Ebury backdoor.
Two camps: those questioning why anyone SSHs over Tor and those defending it for anonymity or NAT traversal.
378 pts · 97 comments
Google open-sourced TorchTPU, a native PyTorch backend for its TPU hardware.
The release is seen as a push to sell TPU hardware beyond Google Cloud, replacing pytorch/xla.
365 pts · 79 comments
A researcher seeded 50,000 fake stats online, and ChatGPT still cites them 7 months later.
Alarm over LLMs' inability to distinguish real from fake data, amplified by citation laundering.
350 pts · 200 comments