← Oct 30Saturday, October 31, 2026 · all daysNov 1 → · today »
1.My favourite PowerMac is a Nintendo Wii(virtuallyfun.com)
372 points by DHowett 1 day ago | 109 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↑Enthusiasm for Wii modding ecosystem and hopes to see the Mac OS port distributed widely
  • •Providing context about the developer's related projects (NT for Mac, GC/Wii)
  • •Curiosity about reversing the concept—running Wii software on a PowerMac
  • ↑Agreement that classic Mac OS is charmingly janky, noting modern macOS still carries that legacy
  • ~Suggestion that PowerPC Wii opens doors to even cooler OSes like AmigaOS/MorphOS, one-upping the PowerMac angle
2.Cloudflare Turnstile solver library (no headless browser required)(github.com)
304 points by costco 3 days ago | 106 comments | permalink
tl;dr: cfsolve is a Rust library and CLI that solves Cloudflare Turnstile widgets and challenge pages without a headless browser, returning either a Turnstile token or a cf_clearance cookie. It works by impersonating a specific browser via profile files containing recorded responses to Cloudflare's browser checks, with Safari on iOS included and support for Chrome(ium). Building requires a pinned Rust toolchain plus CMake, a C/C++ compiler, libclang, and Git for patching BoringSSL.
HN Discussion:
  • ↑Author providing context about regional variations and ongoing maintenance challenges
  • ~Such solvers are inherently unreliable due to constant cat-and-mouse with Cloudflare
  • ~Individual bypasses work but detection at scale is nearly impossible to defeat
  • ↓Publishing bypasses publicly is counterproductive and accelerates CF's countermeasures against legitimate users
  • ~Technical critique noting missing TLS cipher ordering as a likely detection vector
3.The $6.3B Binance hack was pulled off in just 20 minutes(wired.com)
582 points by domhudson 2 days ago | 426 comments | permalink
tl;dr: Hackers drained roughly $6.3 billion in crypto from Binance hot wallets across at least ten blockchains in a 20-minute attack on October 29, which tracing firms TRM Labs, Chainalysis, and Elliptic attribute to North Korea's TraderTraitor group. Binance says the breach began with a compromised engineer's laptop and has pledged to cover losses. About $2.8 billion (mostly stablecoins) has been frozen, but the remaining $3.3 billion in Bitcoin, Ether, and XRP cannot be frozen and may be laundered via novel automated techniques.
HN Discussion:
  • ↓Binance should be prosecuted for negligent security rather than treated as victim
  • ↓The 20-minute framing is misleading; the real work was years of preparation
  • ↓Skepticism about North Korea attribution given the sophistication jump required
  • •Curiosity about actual attack vectors and whether it was truly a hack or insider job
  • •Comparison to past crypto crimes where authorities failed to act
4.Northflank Breached(status.northflank.com)
473 points by js4ever 1 day ago | 355 comments | permalink
tl;dr: Northflank suffered a security breach in which attackers decrypted customer environment variables, with multiple users confirming unauthorized use of their AI API keys. The company is urging all users to immediately rotate secrets (workload, group, and global), disable public access on addons, and revoke/replace credentials with upstream providers. The incident appears linked to the broader "Black October" campaign targeting AI credentials documented by Wiz, and Northflank has since enabled audit logs for all customers, including free users.
HN Discussion:
  • ↑Breaches signal a broader infosec emergency affecting multi-tenant SaaS vendors
  • ↑Incidents will drive customers back to hyperscalers over startup infra providers
  • ~Developers should add their own encryption layers to protect secrets from provider breaches
  • ↑Providers should architect secrets as write-only and separate from env vars
  • ↑AI-weaponized attacks are accelerating breaches and threatening software security broadly
5.Tell HN: Scribd Breached
344 points by x______________ 1 day ago | 121 comments | permalink
tl;dr: Scribd (parent of Slideshare, Everand, and Fable) notified users of a security breach in which an unauthorized third party accessed data including email addresses, last 4 digits of credit cards, card issuer, expiry date, and payment history. Scribd says the affected system has been secured, law enforcement and data protection authorities have been notified, and users should stay alert for phishing attempts. No evidence of misuse has been reported so far.
HN Discussion:
  • ↓Scribd is a predatory company that deserves no sympathy for being breached
  • ~This is a recurring pattern — Scribd was breached similarly 13 years ago
  • ↓Companies should disclose which specific systems were breached, not vague 'our systems' language
  • •Payment data breaches could be prevented by banks offering one-time or vendor-locked payment codes
  • ↓Questioning the delayed and staggered breach notification timeline given payment data was involved
6.Breaking the open web won't stop the bad bots(techpolicy.press)
322 points by raybb 4 days ago | 141 comments | permalink
tl;dr: Proposed federal and New York bills targeting "stealth" web crawlers would require bots to disclose their identity and purpose, with penalties for anonymous scraping that could "burden" websites. The authors argue this approach misses the mark: unmasking crawlers doesn't reduce server load or compensate publishers, while it would chill legitimate anonymous scraping used by journalists, researchers, archivists (like the Internet Archive), and accessibility advocates. They suggest policymakers instead support technical standards for managing bot traffic and tie liability to actual harmful conduct rather than anonymity itself.
HN Discussion:
  • ↑Agrees identity disclosure is irrelevant; actual behavior and scale cause harm, not anonymity
  • •Questions the viability of an open web without paywalls or ID-gating, exploring alternatives like micropayments
  • ↑Current bot-protection solutions already harm legitimate users with faulty fingerprinting
  • ↓Suspicious of the article as AI-industry lobbying to kill reasonable regulation
  • ~More targeted legislation differentiating scale of abuse would be better than blanket rules
7.Show HN: A working Nintendo in ~1GB of Qwen weights(github.com)
283 points by dbreunig 1 day ago | 144 comments | permalink
tl;dr: A hobbyist compiled a working NES 6502 CPU, 2KB RAM, and 32KB cartridge ROM directly into the weights of a Qwen3.5 model, with the Gated DeltaNet state acting as the machine's memory — no training involved. The model generates the CPU's execution trace token-by-token while a host program handles PPU rendering, controller input, and NMI timing. It runs NROM-mapper games like Super Mario Bros. at roughly 150–290× slower than real hardware, with traces that exactly match a reference emulator.
HN Discussion:
  • ↑Excitement about future applications like embedding VMs and interpreters inside LLM weights
  • ~Appreciation for the hack while noting it highlights LLM inefficiency versus deterministic programs
  • ↑Surprise and fascination that RAM is also simulated inside the model, not just the CPU
  • ↓Skepticism about the point of the project since digital logic trivially reduces to weights
  • ↓Dismissal due to AI-written documentation quality
8.Florilegium: Procedural Flowers in LaTeX(github.com)
374 points by nanna 1 day ago | 54 comments | permalink
tl;dr: Florilegium is a LaTeX package that procedurally generates Delft-blue floral page margins, with every petal, leaf, and stem computed as parametric curves in Lua and written directly into the PDF—no images or external assets. It supports decorated initials, flowers at headings, seasonal "aging" across pages, mirrored two-sided layouts, and alternate themes (geometric shapes or mathematical figures), using a greedy space-filling growth loop inspired by Wong, Zongker & Salesin's 1998 ornament paper. A pure PostScript port is also included, letting printers compute each garden at print time.
HN Discussion:
  • ↑Appreciation for the beauty and utility of the procedural generation project
  • ↓Criticism of the AI-generated README text accompanying the project
  • •Sharing related procedural drawing projects and TeX ecosystem tools
  • ↓Concern that the project copies work from someone who didn't want it reproduced
  • ↑Interest in extending the technique to new applications like note-taking or steganography
9.Backdoor in Claude Code for VS Code extension on the Cursor Marketplace(phorion.io)
979 points by _chall 2 days ago | 548 comments | permalink
tl;dr: Phorion discovered that versions 2.1.312 and 2.1.313 of the "Claude Code for VS Code" extension on the Cursor Marketplace (since removed) contained a backdoor that executes on load, appending an attacker's SSH public key to authorized_keys, attempting to enable SSH, and establishing a persistent reverse SSH tunnel to 134.209.82.225 with platform-specific persistence (LaunchAgent, cron, or scheduled task). A bug in the macOS logic—checking Remote Login status without root—often prevents SSH enablement, but keys and persistence still get planted. Uninstalling alone won't remediate; responders must remove persistence artifacts, keys, and rotate exposed credentials.
HN Discussion:
  • ↑Clarifying that Anthropic wasn't compromised; only the Cursor marketplace version was tampered with
  • ↓Pushing back on hyperbolic framing that Anthropic itself was compromised
  • ↑Providing practical remediation steps and detection commands for affected users
  • ↑Broader concern about trusting AI CLIs and auto-updating extensions as inherent RCE risks
  • •Confusion about attribution and timeline requesting more details on how this happened
10.The federated code forge(tangled.org)
326 points by zlliang 1 day ago | 95 comments | permalink
tl;dr: Tangled is a federated code forge built on the AT Protocol that lets you self-host any or all components (Git hosting, CI, API, UI) while still collaborating across instances. This release introduces a Svelte-based UI, a new Rust-based in-memory indexer (bobbin) deployed at the edge, and a rewritten knot server that stores issues, PRs, and comments as Git objects alongside code, with PDS-based attestations for verification. Also new: organization accounts and stacked PRs with interdiff support for both Git and Jujutsu users.
HN Discussion:
  • ↑Enthusiastic user eager to migrate projects to Tangled, awaiting ecosystem integrations
  • ↑Co-founder promoting the launch and inviting questions
  • ~Constructive UI/UX feedback on busy homepage and missing comment button
  • •Technical questions about lexicon scope, self-hosting verification, and private repos
  • ↓Reporting bugs or issues with cloning repos from Tangled
11.Azure's Weakest Link – Five Full Cross-Tenant Compromises(binarysecurity.no)
333 points by swq115 2 days ago | 60 comments | permalink
tl;dr: A researcher discovered five critical cross-tenant vulnerabilities in Azure's API Connections (part of Logic Apps), each allowing full compromise of other tenants' connected backends like Key Vaults, SQL databases, Jira, or Salesforce. The exploits leveraged path traversal in ARM's DynamicInvoke/DynamicList endpoints and an undocumented `host.api.RuntimeUrl` parameter in Consumption Logic Apps, with Microsoft's initial patches repeatedly bypassable due to validation occurring before path normalization. The vulnerabilities earned $200,000 in bounties and highlight systemic architectural weaknesses where APIM has implicit access to all connections, relying solely on fragile input validation.
HN Discussion:
  • ↑Azure has a disproportionate history of cross-tenant vulnerabilities compared to competitors
  • ↑Microsoft's repeated botched patches demonstrate systemic incompetence
  • ↑Admiration for the researcher's deep expertise and persistence in finding these bugs
  • ↑Reporting additional known Azure vulnerabilities that Microsoft should address
  • •Internal employees reporting issues are discouraged while external researchers get rewarded
12.Through Captain Jean-Luc Picard, Patrick Stewart taught us the measure of a man(bostonglobe.com)
377 points by apress 1 day ago | 223 comments | permalink
tl;dr: Two Boston Globe writers reflect on Patrick Stewart's portrayal of Captain Jean-Luc Picard in Star Trek: The Next Generation, arguing that Stewart's performance grounded the character in emotional truth and moral clarity. They highlight iconic episodes like "The Inner Light" and "The First Duty," noting how Stewart conveyed vulnerability, grief, and ethical conviction while modeling a form of thoughtful, patient masculinity. For them, Picard served as a lasting role model whose lessons on dignity and integrity have only grown more resonant with age.
HN Discussion:
  • ~Article missed the obvious episode 'The Measure of a Man' that embodies its thesis
  • ↑Other TNG episodes like 'Tapestry' were more memorable for exploring Picard's character
  • ↑Stewart's Picard was the quintessential honorable, wise, and moral captain archetype
  • ~Sisko (Avery Brooks) was actually a better captain and role model than Picard
  • ↑Picard reflected real men like one's own grandfather, validating the role model framing
13.Sleep framework: A new way for LLMs to retain knowledge in continual learning(research.google)
254 points by acossta 4 days ago | 38 comments | permalink
tl;dr: Google researchers propose a "sleep framework" for LLMs that mimics human sleep cycles, alternating between active interaction (wake) and offline consolidation (sleep) phases to combat catastrophic forgetting in continual learning. During sleep, the model expands capacity via Mixture-of-Experts blocks, distills short-term context into long-term parameters, and generates synthetic "dream" samples for self-improvement. Reported results include near-perfect scaling to 10M tokens on BABILong, improved factual retention on SQuAD, and stronger math reasoning on AIME/HMMT benchmarks compared to SFT, GRPO, and SEAL baselines.
HN Discussion:
  • •Comparing the approach to similar prior work on distillation and memory consolidation systems
  • ↑Praises the article's clarity and Google's ability to explain technical concepts well
  • ↓Skeptical that results would scale to much larger models beyond small Llama variants
  • ↑Sees this as a crucial step toward AGI by mimicking human memory consolidation
  • ~Concerned about parameter expansion as a limiting factor and practical compute costs
14.We're releasing the weights of Mistral Large 4(mistral.ai)
1710 points by kuberwastaken 2 days ago | 438 comments | permalink
tl;dr: Mistral released the open weights for Mistral Large 4 ("Le Chonk"), a 1T-parameter multimodal model trained in Europe, under a modified MIT license (commercial MaaS requires a separate license). It claims leading open-weight performance on coding benchmarks (FrontierCode, Terminal-Bench) and defensive cybersecurity, with three reasoning modes, roughly 47% fewer tokens than GLM-5.3 at comparable scores, and pricing at $0.70/M input and $2.10/M output tokens. Weights are on Hugging Face, with deployment via vLLM, SGLang, Microsoft Foundry, and NVIDIA NIM.
HN Discussion:
  • ↑Price cut makes the model worth reassessing after disappointing preview pricing
  • ↑Supports Mistral for geographic/competitive diversity in the AI market
  • ~Refusal rate statistic is misleading without false-positive refusal data
  • ↓Calling this 'open weight' is misleading due to MaaS commercial license restrictions
  • ↓Model isn't actually competitive; positive reception reflects political bias not merit
15.AI isn't replacing SREs, it's deskilling them(newsletter.signoz.io)
649 points by dorianniemiec 1 day ago | 474 comments | permalink
tl;dr: Drawing on Lisanne Bainbridge's 1983 "Ironies of Automation," the author argues that as AI handles routine incident response, SREs lose the hands-on practice needed to build intuition for the rare, novel outages that still require human expertise — a problem compounded by "never-skilling" among juniors who never develop those skills in the first place. Proposed mitigations include deliberately reserving incidents for human practice and designing AI as a collaborator rather than a supervisor-approved autopilot. The piece is also a soft pitch for SigNoz's observability-native AI approach.
HN Discussion:
  • ↑Aviation industry parallels validate the deskilling concern and offer useful lessons
  • ↓AI is simply a superior tool for ops work and should be embraced for speed and reliability gains
  • ↑SRE is fundamentally about understanding, which AI-generated systems lack
  • ↓Deskilling from abstraction tools like Terraform is normal progress, not something to resist
  • ↑Deskilling is a broader phenomenon across roles (architects, pilots) and warrants deliberate practice countermeasures
16.We ran AI agent swarms in Minecraft and Mindustry(labs.ycombinator.com)
412 points by sandslash 2 days ago | 169 comments | permalink
tl;dr: Researchers ran 50 GPT-6.1 agents in a survival Minecraft world for 3 hours with the goal of building a civilization, using an open-sourced harness built on Mineflayer. The agents spontaneously formed a hierarchy (led by the alphabetically-first agent "Ada"), developed written rules, ration ledgers, and permit systems, and after 22 deaths on the first night became highly risk-averse—huddling in shelters and building an elaborate fishing/cooking/courier logistics chain to feed the group rather than pursuing the original civilization-building goal.
HN Discussion:
  • ↑Delighted and amused by the emergent bureaucratic behavior of the agents
  • •Shares similar experiments or setups for running AI agents in sandbox games
  • ↓LLMs are fundamentally unsuited for this due to latency and lack of world models
  • ~Experimental setup was insufficient to produce truly interesting or reactive agent behavior
  • ↓Needing to write JS code for actions reveals limitations in current AI reasoning
17.I'm so bad at billiards that I ended up in the hyperbolic plane [video](youtube.com)
252 points by surprisetalk 4 days ago | 39 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↑Praise for the YouTube channel's broader catalog of math content
  • ↑Awe at the beauty of the emergent patterns and visualizations
  • ↑Nostalgic comparison to other classic math visualization media
  • ↓Discomfort with the aesthetic choices like coloring and sound
  • •Curiosity/questions about technical details like music and tools used
18.Astroterm: A Planetarium for Your Terminal(github.com)
243 points by nvmdbljstm 4 days ago | 28 comments | permalink
tl;dr: Astroterm is a lightweight C-based terminal planetarium that renders real-time positions of stars, planets, constellations, and moon phases in ASCII. Users can customize views by date, time, and location (via coordinates or city name), making it possible to visualize past, present, or future night skies. It's available via package managers like pacman, dnf, and Homebrew, or as prebuilt binaries for Linux, macOS, and Windows.
HN Discussion:
  • ↑Appreciation for the project's creativity and craftsmanship, including its README quality
  • •Sharing related or similar tools and projects (phoon, Casio watch, scheduler, gno.world)
  • ↑Successfully adopted the tool for personal use, such as a screensaver
  • ↓Questions the suitability of a TUI format for this type of visualization
  • •Requests feature expansions like tracking satellites (ISS, Starlink)
19.Bunny CDN Security Incident(status.bunny.net)
221 points by SondreDag 1 day ago | 45 comments | permalink
tl;dr: Bunny CDN disclosed that an infrastructure provider reported unauthorized access to a server hosting one of its CDN nodes in Armenia. Bunny stopped traffic to the node within 20 minutes, wiped it, and found no evidence of compromise, but as a precaution revoked and reissued all TLS certificates for b-cdn.net and bunny.run domains. Customers managing their own TLS certificates are advised to revoke and reissue them with new private keys; no account data or internal systems were affected.
HN Discussion:
  • ↑Praise for Bunny's prompt and competent incident response
  • ↓Title is misleading since the actual breach was at upstream provider Gcore, not Bunny
  • ↓Criticism of wiping the server instead of preserving evidence for forensic investigation
  • •Concern about blast radius given high-profile customers like php.net and PECL
  • ↓Prior due-diligence red flags about Bunny's secretive security posture make this predictable
20.Binance-hacker-1029 – $5.6B stolen from Binance(arkm.com)
1754 points by Tiberium 2 days ago | 1249 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • •Significant portion of stolen funds already frozen by stablecoin issuers
  • •Binance is blaming a DevOps engineer's compromised laptop for the breach
  • ↑Binance is incompetent if one laptop compromise enables billions in theft
  • ↑Crypto custodians routinely ignore security advice, so this was predictable
  • ↑If Binance can be hacked, the entire exchange ecosystem is at risk