Jun 15Tuesday, June 16, 2026 · all daysJun 17 · today »
1.The time the x86 emulator team found code so bad they fixed it during emulation(devblogs.microsoft.com)
492 points by paulmooreparks 1 day ago | 168 comments | permalink
tl;dr: A Windows x86-32 binary translation emulator team encountered a program that allocated 64KB on the stack and initialized it by unrolling the init loop into 65,536 individual byte-write instructions—256KB of code to zero 64KB of data. Rather than faithfully translating this monstrosity, the team added special-case detection to recognize the pattern and replace it with an equivalent tight loop during translation.
HN Discussion:
  • Sharing similar war stories of inefficient code discovered while working on tooling or emulation layers
  • Compatibility/translation layers fixing broken software is a recurring and valuable pattern
  • Questioning the article's framing—loop unrolling may actually be a legitimate optimization choice
  • ~Correcting or nuancing the article's claim about the 'standard way' of stack allocation
  • Broader lamentation about pervasive unoptimized code in modern software
2.A backdoor in a LinkedIn job offer(roman.pt)
1573 points by lwhsiao 1 day ago | 301 comments | permalink
tl;dr: A developer received a LinkedIn message from a fake crypto startup "recruiter" asking him to review a GitHub repo and check a "deprecated Node modules issue" — bait to trigger `npm install`, which auto-runs a `prepare` script that executes a backdoor disguised as a test file, fetching and running arbitrary code from a remote server. Both the recruiter's LinkedIn profile and the repo's commit author identity were stolen from real people. He flagged the threat using a read-only AI agent on a throwaway VPS, which spotted the payload in seconds.
HN Discussion:
  • Shares similar scam experiences confirming this attack pattern is widespread
  • Calls for organized societal/law enforcement response to cybercrime
  • Criticizes LinkedIn and GitHub (Microsoft) for inaction on reported threats
  • Notes the attack mimics legitimate interview tasks, making it dangerously deceptive
  • Suspects the writeup was largely AI-generated, casting doubt on authenticity
3.I admire Fabrice Bellard. He is almost certainly a better overall programmer(twitter.com)
905 points by apitman 1 day ago | 443 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Bellard's genius lies in choosing impactful problems, not just technical skill
  • ~Bellard's strength is translating complex specs into efficient C implementations
  • Bellard's legacy in FFmpeg is overstated; his original code was poor and long replaced
  • Admiration for Bellard's reclusive, focused work ethic and avoidance of limelight
  • The article's claim that FFmpeg is the Internet's invisible engine is hyperbole
4.Iroh 1.0(iroh.computer)
1365 points by chadfowler 2 days ago | 446 comments | permalink
tl;dr: Iroh 1.0 is the first stable release of a networking library that replaces IP-based addressing with public keys, enabling secure, direct device-to-device connections that persist across network changes and NAT boundaries. The release locks in wire protocol and API stability, adds official bindings for Python, Node.js, Swift, and Kotlin alongside Rust, and includes QUIC multipath, NAT traversal, WASM support, and pluggable transports like BLE and Tor. The project's public relays have handled 200M+ endpoints in the last 30 days, with ~95% of connection traffic typically flowing directly peer-to-peer.
HN Discussion:
  • Iroh is best understood as Tailscale at the application layer for app developers
  • Production users praise Iroh's reliability, team responsiveness, and developer experience
  • Decentralized peer-to-peer networking is the future and Iroh advances that vision
  • The announcement lacks clarity on fundamentals like what 'keys' mean and how relays work
  • Skepticism that Iroh solves a real problem given IP, DNS, IPv6, and QUIC already exist
5.Banned Book Library in a Wi-Fi Smart Light Bulb(richardosgood.com)
570 points by sohkamyung 1 day ago | 347 comments | permalink
tl;dr: A hacker reflashed a cheap Tasmota-based WiFi smart bulb (ESP32-C3, 4MB flash) to host an open WiFi access point serving banned ebooks via a captive portal, intended as a discreet "digital dead drop" you could install in public. Key challenges included the tiny storage (solved by repartitioning flash to free up 2MB for books), writing a custom safeboot firmware to enable OTA updates without leaking WiFi credentials, and abandoning a microSD expansion idea because soldering destroyed the bulb. Future plans include RGB color matching and mesh networking between bulbs to share larger libraries.
HN Discussion:
  • Free flow of information is essential to resist tyranny, supporting the project's intent
  • Notes this concept is not new, referencing prior PirateBox/LibraryBox projects
  • Skeptical that the light bulb form factor actually provides meaningful stealth or undetectability
  • ~Practical concerns about usability, like Android auto-disconnecting from internet-less WiFi
  • Criticism of the banned book selection as mainstream and lacking real intellectual diversity
6.Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?
1238 points by cloudking 2 days ago | 532 comments | permalink
tl;dr: A Hacker News user asks whether anyone has fully replaced Claude or GPT with a local model for their primary coding workflow, rather than just experimental use. They're requesting details on setups and performance metrics like tokens per second.
HN Discussion:
  • Successfully replaced cloud models with local Qwen/Gemma setups for privacy and cost savings
  • ~Local models work for most tasks but still fall back to frontier models for complex work
  • Local models aren't worth the effort given the opportunity cost versus frontier models
  • ~Open source models via third-party inference providers offer better speed/cost than local or big providers
  • Expectations matter; appropriately-sized local models are good enough for scoped tasks
7.TinyWind: A pixel pirate sailing game with real wind physics (380k+ kms sailed)(tinywind.io)
984 points by tinywind 1 day ago | 182 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Game is impressive and enjoyable but needs UI/control clarifications and feedback
  • Wind physics aren't truly realistic; sailing mechanics like tacking and dead angles are missing
  • ~Difficulty is too high with perfect enemy aim and limited healing options
  • Suggestions for added realism like rudder-only-with-speed and richer simulation features
  • ~Browser registration model creates friction; prefer downloadable standalone version
8.I Love the Computer(michaelenger.com)
298 points by speckx 1 day ago | 158 comments | permalink
tl;dr: A developer reflects on a lifelong love affair with computers, tracing his journey from a childhood IBM 486 in 1990s Norway through print magazines, early Internet exploration, and a career in programming. Inspired by a podcast quote from Chris Person, he laments how AI hype, dark patterns, and "brogrammer" culture have corrupted the once-idealistic tech space, but remains hopeful about decentralized alternatives and is content to retreat back into being a niche enthusiast.
HN Discussion:
  • Still loves the computer itself but finds the surrounding industry hard to like
  • Disagrees with framing AI as snake oil since LLMs are genuinely useful for learning
  • ~Nostalgia is partly age-related romanticizing; the old internet still exists for those who seek it
  • Shares personal resonance with the author's childhood computing memories and emotional connection
  • Catalogs specific moments when computing's promise was corrupted by corporate control and DRM
9.Hetzner Price Adjustment(docs.hetzner.com)
527 points by tuhtah 2 days ago | 733 comments | permalink
tl;dr: Hetzner is raising prices on dedicated servers and cloud instances starting June 15, 2026, applying to new orders and cloud rescales (existing orders keep old pricing). Cloud server increases are particularly steep, with CCX (dedicated vCPU) instances roughly doubling or tripling in price—e.g., CCX13 jumps from €15.99 to €42.99/month, and CCX63 from €374.49 to €853.49/month. CPX and CAX ARM instances also see significant hikes, while standard CX shared instances rise more modestly (~25-40%).
HN Discussion:
  • Price increases reflect real hardware cost rises in RAM and disk
  • The 2-3x hike is excessive and unjustified, unlike a reasonable 25-50% bump
  • Users are looking to migrate to cheaper alternative providers
  • Hetzner was undercharging and is finally capturing market value
  • ~Hetzner has stagnated on hardware refreshes, reducing their value proposition
10.My Homelab AI Dev Platform(rsgm.dev)
357 points by rsgm 2 days ago | 56 comments | permalink
tl;dr: The author built a homelab AI dev platform by running OpenCode's web UI as a systemd service on an isolated VM, giving it a dedicated Git user that can push branches but not merge to deploy. Changes flow through PR review and GitOps (Arcane for Docker stacks) for deployment, enabling infra updates from any device—including a phone—without granting AI direct access to production services. The main gap is CI feedback, since Forgejo Actions doesn't expose job logs via its public API.
HN Discussion:
  • Commenters share their own similar homelab AI setups, validating the approach
  • ~Alternative implementations using CI runners or polling timers instead of persistent server
  • AI-assisted homelab management has transformed feasibility and is enjoyable
  • The author's site is unreachable or DNS-blocked, undermining credibility
  • Skepticism about hardware cost requirements before even reading
11.Fox to buy Roku(wsj.com)
353 points by thm 2 days ago | 418 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Roku's content involvement and ads have alienated long-time customers, this acquisition confirms fears
  • A media giant like Fox should not be allowed to control TV hardware in so many households
  • Fox's reputation and business practices make this acquisition a reason to abandon Roku devices
  • Recommendations to switch to alternative hardware like Nvidia Shield or Wyze cameras
  • Clarifying which Fox entity is involved and providing alternative news sources
12.Copper transport drug restores memory and clears toxic Alzheimer's proteins(monash.edu)
349 points by bookofjoe 2 days ago | 139 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Amyloid-targeting therapies have repeatedly failed, suggesting the article's premise is flawed
  • Results are only in mice and headline overstates implications for humans
  • ~Practical concerns about drug dosage and liver toxicity at required levels
  • Cautious optimism that the mechanism could be valid even if amyloid isn't the root cause
  • ~Personal/anecdotal context noting Alzheimer's complexity and subtypes limit single-treatment hopes
13.Salesforce to Acquire Fin (formerly Intercom) for $3.6B(salesforce.com)
323 points by colesantiago 2 days ago | 239 comments | permalink
tl;dr: Salesforce is acquiring Fin (formerly Intercom) for approximately $3.6 billion to bolster its Agentforce platform with Fin's AI customer service agent, which runs on a proprietary model called Apex and reportedly resolves 76% of support volume end-to-end. The deal targets faster time-to-value for SMB and commercial customers and brings Fin's 30,000+ customer base to Salesforce, with closing expected in Q4 of Salesforce's fiscal 2027.
HN Discussion:
  • AI customer service can be excellent when executed well, supporting Fin's value proposition
  • AI support agents are frustrating and provide negative value compared to humans
  • Intercom/Fin has little moat as businesses can build their own AI support agents cheaply
  • Acquisition is strategic competitive move by Benioff against Sierra and to control CRM integration
  • ~$3.6B valuation seems surprisingly low given Fin's market position and undermines broader AI valuations
14.Game Engine White Papers: Commander Keen(forgottenbytes.net)
231 points by mfiguiere 1 day ago | 79 comments | permalink
tl;dr: A 214-page full-color technical book dissecting the original 1990 Commander Keen game engine is now available, free as a PDF or in paperback on Amazon, with source code on GitHub. The book details the hardware (80286, EGA, sound cards), game assets, engine internals, and the CGA port, based on three years of reverse-engineering and experimentation.
HN Discussion:
  • Author used ChatGPT to rewrite and copied content from another author without attribution
  • ~Appreciates the book but suggests adding context comparing PC vs SNES sprite rendering hardware
  • Recommends similar reverse-engineering resources like Cosmodoc and Masters of Doom
  • Requests expanded coverage of other Apogee/Epic era games
  • Requests additional formats like epub beyond PDF
15.US battery manufacturing output continues to break records(fred.stlouisfed.org)
225 points by epistasis 1 day ago | 193 comments | permalink
tl;dr: US battery manufacturing output, as tracked by the Federal Reserve's Industrial Production index (NAICS 33591), continues to hit record highs. The seasonally adjusted index, based at 2017=100, reflects sustained growth in domestic battery production capacity and real output.
HN Discussion:
  • ~US production is growing but tiny compared to China's massive lead
  • Growth is good news for national security and domestic capacity
  • The data is misleading because it likely includes primary/disposable batteries, not EV-relevant production
  • The chart and headline are unclear or editorialized, lacking meaningful context
  • Curious about why China dominates and how the US could replicate it (e.g., nationalizing Chinese plants)
16.Openrouter Fusion API(openrouter.ai)
214 points by tdchaitanya 2 days ago | 84 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Multi-model judging doesn't actually produce better answers, just self-similar ones
  • ~Fusion is too slow and expensive for general use, situational only
  • Sharing personal projects/tools inspired by or similar to Fusion
  • Anecdotal testing showed Fusion produced worse results than alternatives
  • Skeptical questioning whether model variance is real or just temperature randomness
17.Improvement in advanced Alzheimer’s disease following high-dose psilocybin(frontiersin.org)
212 points by cl3misch 2 days ago | 150 comments | permalink
tl;dr: A case report describes an octogenarian woman with 10-year advanced Alzheimer's who, after a 5g dose of psilocybin mushrooms, showed transient multidomain improvements including restored urinary continence, independent ambulation, autobiographical speech, and social engagement—gains persisting for weeks and reinforced by a second 3g session. The authors stress this isn't disease reversal but suggests residual functional capacity may remain accessible via neuromodulation; limitations include the single-case design, no biomarker confirmation, no standardized cognitive scales, and inability to rule out spontaneous fluctuation.
HN Discussion:
  • Methodological skepticism about case report quality, predatory journal, and unsupported claims
  • Ethical concerns about dosing a vulnerable patient outside research protocols
  • Related phenomena like terminal lucidity suggest the findings may be plausible or explainable
  • Excitement about psilocybin's therapeutic potential and need for more research
  • Title is editorialized/overstated relative to what a single case report can show
18.Typst 0.15.0(typst.app)
327 points by schu 1 day ago | 89 comments | permalink
tl;dr: Typst 0.15.0 adds variable font support, MathML-based equations in HTML export, multi-file "bundle" export, multiple bibliographies per document, spot colors, and a new file path type. Breaking changes include stricter HTML paragraph grouping, revised baseline handling for boxes/blocks/lists, removal of previously deprecated elements (path, pattern, pdf.embed), and trimming of "Variable"/"VF" suffixes from font family names. A new `within` selector, `divider` element, and `typst eval` CLI subcommand (replacing `typst query`) round out the release, along with fixes for list marker alignment and CJK justification.
HN Discussion:
  • Typst excels for book/document production despite some LLM tooling friction
  • Specific new features like multiple bibliographies and path type solve real workflow problems
  • ~Footnote handling remains a significant unresolved weakness for academic writing
  • ~Typst's math mode syntax dropping backslashes was a design mistake
  • Typst is a superior alternative to LaTeX/Word for technical documents
19.CrankGPT(crankgpt.com)
598 points by rishikeshs 2 days ago | 233 comments | permalink
tl;dr: CrankGPT is a satirical "human-powered" AI product pitch that mocks the AI industry's energy consumption, privacy issues, and wealth concentration. The premise: a hand-cranked (or pedal-powered for "power users") device that generates tokens locally via physical effort, addressing climate concerns, data privacy, tech CEO enrichment, and fitness all at once. It's a tongue-in-cheek critique framed as a product page, complete with "request a demo" call-to-action.
HN Discussion:
  • ~Website design is poor; technical documentation page is much better reading
  • Enthusiastically embraces the concept as personally appealing and would actually want one
  • Designing under hand-crank energy constraints is a genuinely valuable engineering framing
  • Anti-AI energy protest is confused; green energy already solves this without manual labor
  • Satire is indistinguishable from reality, echoing Matrix-style human-battery dystopia
20.Your ePub Is fine(andreklein.net)
895 points by sohkamyung 2 days ago | 305 comments | permalink
tl;dr: An author's EPUB passed epubcheck validation and worked on Kindle, Apple Books, and Thorium, but failed silently on Kobo devices. The culprit was a single line of modern CSS (`max-width: min(150px, 30vw)`) that Adobe's RMSDK rendering engine—frozen circa 2013 and used by Kobo for standard EPUBs—couldn't parse, causing it to drop the entire book instead of ignoring the unknown declaration as the CSS spec has required since 1996. Workaround: rename files to `.kepub.epub` to route them to Kobo's actively maintained WebKit renderer instead.
HN Discussion:
  • Adobe has a long history of poor software maintenance and abandoning ecosystems
  • RMSDK is inaccessible and Adobe is unresponsive to developers
  • EPUB spec itself is problematic and epubcheck isn't as reliable as the author suggests
  • Kobo is moving away from RMSDK; workarounds like kepubify or .kepub.epub help
  • ~EPUB readers vary so much that minimal styling is the safest approach