Aug 18Wednesday, August 19, 2026 · all daysAug 20 · today »
1.OpenLogi(openlogi.org)
1608 points by amatheus 3 days ago | 425 comments | permalink
tl;dr: OpenLogi is a local-first, Rust-based alternative to Logitech Options+ that talks directly to Logitech mice over HID++ to remap buttons, control DPI, and configure SmartShift, with no account or telemetry. Bindings live in a plain TOML config file, and it supports Bolt, Unifying, Lightspeed, Bluetooth, and wired connections across signed macOS, Linux, and Windows builds. Per-app profiles, Options+ import, and Flow-style cross-machine features are on the roadmap; users must quit Options+ (or Solaar on Linux) to avoid HID++ conflicts.
HN Discussion:
  • Logitech's official software is bad and needs open source alternatives
  • Pointing to existing alternatives like Solaar, BetterMouse, or BetterTouchTool
  • AI/LLM-generated website copy and vibe-coded projects undermine trust in the tool
  • AI is empowering developers to replace poor vendor software, sharing similar reverse engineering efforts
  • Hardware vendors should adopt better approaches like WebUSB or offline offerings
2.Cerebras CS-4(cerebras.ai)
453 points by sunils34 3 days ago | 264 comments | permalink
tl;dr: Cerebras announced the CS-4, a rack-scale AI system packing three WSE-3 Turbo wafers that claims up to 30x faster inference than GPUs and 10x better throughput per watt than the prior CS-3. Key architectural changes include a modular "backpack" design with power delivery just 0.5mm from the processor, a new programmable I/O subsystem enabling 2-microsecond wafer-to-wafer latency, and support for 1,000+ tokens/sec on 10T+ parameter models. First shipments begin this quarter.
HN Discussion:
  • Cerebras neglects developer/API offerings by only providing outdated models despite hardware claims
  • ~Underwhelmed that this is a WSE-3 refresh rather than the expected WSE-4
  • The article inadvertently reveals GPT-5 model parameter counts through its specs
  • Skepticism about real-world performance given predecessor's lackluster market adoption
  • Optimism that Cerebras and others will meaningfully challenge NVIDIA's monopoly
3.Being ambitious and being a dad(nicholascharriere.com)
905 points by nichochar 5 days ago | 707 comments | permalink
tl;dr: A founder reflects on the tension between startup ambition and fatherhood, noting that many admired builders (Jobs, Einstein, Musk) were notoriously bad parents, and rejecting the common trade-off of delegating parenting. He disputes the "quality time" theory in favor of maximizing quantity, especially when kids are young, and acknowledges his output is measurably lower than his pre-kids self. His approach: redefine ambition to include being a great dad, then compensate through ruthless focus, health, strict family time rules, and cutting time-wasters.
HN Discussion:
  • ~Chose not to have kids to preserve energy for ambitious work, acknowledging the real trade-off
  • Confirms you can be ambitious and a present dad by cutting out time-wasters like TV and friends
  • ~Ambition must include humility and checking in with your partner who shoulders hidden work
  • Applauds the author for calling out founders who offload parenting to partners
  • Extreme career success is a lottery, so prioritizing kids sacrifices little in expectation
4.A 3D fruit fly on macOS desktop powered by the real FlyWire connectome(github.com)
372 points by phoenix120 3 days ago | 182 comments | permalink
tl;dr: A macOS desktop toy that renders a 3D fruit fly driven by a live 1 kHz leaky-integrate-and-fire simulation of a 668-neuron circuit with ~19,000 real synaptic connections from the FlyWire connectome (FAFB v783). Cursor motion feeds real LC4/LPLC2 looming detectors, and escape only triggers when the Giant Fiber actually spikes — no scripted behaviors. Includes an interactive brain window showing 23,210 neuron soma positions where you can "optogenetically" stimulate regions and watch the fly groom, walk, steer, or flee accordingly.
HN Discussion:
  • Project overstates connectome control since behaviors are still scripted and merely triggered
  • Questioning the ethics of simulating a living creature's brain
  • Suggesting integration with existing biomechanical fly simulators to enhance realism
  • Speculating about future potential of booting entire organisms from DNA sequences
  • ~Complaints about AI-generated README and missing demo video
5.Children's stunted lungs show recovery in ultra low emission zone(bbc.com)
409 points by dabinat 3 days ago | 403 comments | permalink
tl;dr: A five-year study of 3,400 children found that London kids' previously stunted lung growth accelerated to match those in less-polluted Luton after the 2019 introduction of the Ultra Low Emission Zone, with clinically impaired lung capacity dropping from 14% to 9%. Researchers linked the improvement to faster reductions in nitrogen dioxide exposure in London, though independent experts caution that Covid-era changes and behavioral shifts (like more walking/cycling) could also be factors. Published in Lancet Public Health, the study is being called the strongest evidence yet that clean air zones can reverse pollution damage in children.
HN Discussion:
  • Personal anecdotes confirming pollution reduction improves respiratory health
  • Broader examples from China and Japan supporting pollution reduction benefits
  • Skeptical of study methodology and confounding factors like COVID lockdowns
  • Findings are obvious and shouldn't require debate given historical precedent
  • Sharing the original study data and technical details neutrally
6.How does IKEA come up with names for its products?(ikea.com)
448 points by NaOH 3 days ago | 319 comments | permalink
tl;dr: IKEA names products with real Swedish words (4-12 letters, ideally containing Å, Ä, or Ö) instead of numbers, a system started by founder Ingvar Kamprad who struggled with numeric codes. Names follow category conventions—sofas get Swedish place names, bookshelves get men's names, children's products get animals/nature—and are vetted to avoid trademarks, surnames, and unwanted meanings in other languages. IKEA names 2,000-3,000 new products this way each year.
HN Discussion:
  • Counter-examples showing IKEA's vetting process fails, with names having unfortunate meanings in other languages
  • ~Sharing amusing accidental wordplay or puns IKEA names create in other languages
  • Confirming the naming conventions described (place names, categories) with specific examples
  • Questioning or scrutinizing specific claims in the article, like product counts or letter limits
  • Personal anecdotes about using Swedish IKEA names in everyday life
7.Finger: the 1971 social network that never died(en.andros.dev)
323 points by andros 3 days ago | 119 comments | permalink
tl;dr: Finger, a 1971 protocol from Stanford's AI Lab, was arguably the first social network: users published a plain-text `.plan` file readable by anyone via `finger user@host` on TCP port 79, famously used by John Carmack as a dev journal and even by a CMU Coke machine (an early IoT device). It fell out of favor after the 1988 Morris worm exploited a `fingerd` buffer overflow, but the protocol still ships on macOS, Linux, and Windows, and is seeing a small-web revival via services like Happy Net Box, Finger.Farm, and plan.cat.
HN Discussion:
  • Personal nostalgia sharing early experiences using finger as social/communication tool
  • ~Finger was just one part of a broader Unix-based social network ecosystem
  • ~Modern infrastructure (NAT, multi-device users) makes finger's original model impractical today
  • Active builders promoting finger revival tools and services
  • ~Related protocols like WebFinger deserve mention in this discussion
8.Turbovec – Google's TurboQuant for vector search in Rust(github.com)
290 points by fittingopposite 3 days ago | 34 comments | permalink
tl;dr: Turbovec is a Rust vector index (with Python bindings) implementing Google Research's TurboQuant quantizer, compressing a 10M-doc corpus from 31GB to 4GB while outperforming FAISS IndexPQFastScan by ~3.4× at 4-bit and ~23% at 2-bit across ARM and x86. It supports online ingest with no training phase, SIMD-accelerated search with allowlist filtering, and crash-safe incremental saves. Drop-in integrations exist for LangChain, LlamaIndex, Haystack, and Agno, targeting local/air-gapped RAG stacks.
HN Discussion:
  • FAISS benchmark comparison is outdated since FAISS is no longer state-of-the-art
  • Excitement about compression ratio and potential for local/lightweight deployments
  • TurboQuant itself has unresolved academic misconduct allegations, tainting this project
  • README appears AI-generated, hurting adoption prospects
  • Alternative tools like Qdrant already integrate TurboQuant, questioning need for this
9.Meta's blockbuster trial draws parallels to big tobacco(economist.com)
308 points by newsomix9xl 3 days ago | 288 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Facebook and tobacco parallels are apt; both knowingly harmed users and targeted children
  • ~Regulating addictive platforms is legally messy and hard to define in general terms
  • Meta's relentless engagement optimization effectively created addiction regardless of explicit intent
  • Addiction is a broader human/societal issue, not solely caused by Meta's products
  • Questioning the legal basis of the case since making platforms addictive isn't clearly illegal
10.And then the men with guns tell you to do it anyway(shkspr.mobi)
440 points by _djo_ 3 days ago | 383 comments | permalink
tl;dr: During Egypt's 2011 revolution, Vodafone was forced to broadcast pro-regime propaganda via SMS after armed authorities compelled compliance, illustrating how emergency alert infrastructure can be weaponized. The author, who worked on the UK's Common Alerting Protocol adoption, questions whether network operators can or should refuse government alert requests, citing a recent UK wildfire alert as a benign example that nonetheless raises abuse concerns. The piece concludes that designing an alert system that is both compulsory and resistant to misuse may be fundamentally impossible.
HN Discussion:
  • Technology cannot solve fundamentally social/political problems, only societies can
  • Trust is the missing foundation that makes such systems workable or not
  • Decentralization or staged rollout with engagement metrics could mitigate abuse
  • The author gives up too easily by accepting a generic PSA as proportionate
  • Real-world examples (South Korea) show alert fatigue as a practical concern
11.The Amazon tax(seths.blog)
1351 points by herbertl 3 days ago | 682 comments | permalink
tl;dr: Amazon earns roughly $1 billion per week from search ads that Seth Godin argues degrade search quality, since the platform already knows the best product for a query and ads mainly serve to divert buyers or force sellers to pay to defend sales they'd have made anyway. He contends these zero-sum ads don't grow demand but raise consumer prices, incentivize shoddier products (as click budgets outweigh brand quality), and give Amazon a perverse incentive to worsen organic results. His conclusion: Amazon's ad model amounts to legalized theft from the customers it claims to serve.
HN Discussion:
  • Same predatory ad behavior exists on Google Play Store and other platforms, reinforcing the critique
  • Legal action like trademark infringement or fraud claims should be pursued against Amazon's ad practices
  • This is just rent-seeking behavior enabled by monopoly power across all Big Tech platforms
  • ~Expecting corporations to serve customers is naive; they exist to serve shareholders
  • Ads serve legitimate purposes like helping new/quality products break through or exposing consumers to alternatives
12.Claude writing a macOS driver for my obscure HP printer built only for Windows(twitter.com)
312 points by porridgeraisin 3 days ago | 219 comments | permalink
tl;dr: Summary not available.
HN Discussion:
  • Similar success stories reverse-engineering proprietary drivers/hardware using Claude and tools like Wireshark
  • Headline is misleading since Claude didn't write a driver, just bridged HP's Linux driver in a VM
  • LLM wasted effort reinventing existing solutions instead of using known Linux-AirPrint projects
  • ~Despite the hacky approach, the pragmatic result of working printing justifies the effort
  • Claude's overly friendly conversational tone feels off-putting or creepy
13.Beware Management Consultants(about.iceland.co.uk)
604 points by KolmogorovComp 3 days ago | 155 comments | permalink
tl;dr: A satirical parable about two rowing teams: the Green team wins with 7 rowers and 1 captain, while the Red team loses with 7 captains and 1 rower. Rather than fix the obvious problem, Red team management hires consultants who recommend restructuring into more layers of leadership, then fire the rower after losing again, reward the managers, and ultimately outsource rowing to India — a critique of how consultants and executives protect hierarchy while blaming workers.
HN Discussion:
  • Shares related example of founder-driven company culture critiquing bureaucracy
  • Self-reflection about being complicit in the very bureaucracy the article critiques
  • Agrees consultants have misaligned incentives and management overvalues change
  • Extends the satire by mocking agile methodology as another layer of consultant-driven nonsense
  • ~Blames corporate leaders/hierarchy-brain rather than just consultants for the dysfunction
14.Sticky wage norms and the real wage cost of unexpected inflation(bfi.uchicago.edu)
384 points by jplusequalt 3 days ago | 269 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • Reinforces findings by highlighting that many workers still suffered real wage cuts despite job hopping
  • Argues inflation's impact is understated, especially in high cost areas or for aspirational goods
  • Criticizes the paper for excluding total compensation components like stock and benefits
  • Personal anecdotes confirming real wage declines after job changes or RSU losses
  • Contextualizes findings by noting timing (post-Covid) or sector-specific dynamics like tech
15.Teaching my kid to code with a modern MUD(tau.dev)
263 points by andrewjanke 9 days ago | 69 comments | permalink
tl;dr: A developer built Canon, a web-based multiplayer MUD, to teach his 8-year-old daughter to code, inspired by HyperCard's accessible "view source" ethos. The game uses a deliberately limited scripting language called Cant—optimized for learnability over efficiency—where players create rooms, items, and NPCs with text, images, and simple state (marks and numbers), and everything is cloneable and hackable by design. It includes an optional daily LLM "Wish" feature for generating starter items, and is currently in invite-only beta.
HN Discussion:
  • Project is wonderful and criticism in the thread is unwarranted
  • MUDs are a nostalgic and effective gateway into programming for kids
  • ~Teaching kids to code works best when the environment engages them, and AI's role remains inconclusive
  • Sharing related alternative approaches to teaching kids coding/game creation
  • Appreciation for the aesthetic and craft (ASCII art, design) of the project
16.Ask HN: GitHub employees what's going on? Why?
306 points by sharts 4 days ago | 252 comments | permalink
tl;dr: An Ask HN post requesting that GitHub employees provide insider perspective on why recent outages and issues keep occurring. The poster notes that most existing discussion is speculation based on public data, and wants firsthand, non-PR information from people who actually see what's happening internally.
HN Discussion:
  • AI-boosted coding has caused massive commit volume growth that GitHub's infrastructure can't handle
  • The troubled Azure migration is the root cause of GitHub's reliability issues
  • GitHub's reliability decline correlates directly with the Microsoft acquisition
  • GitHub leadership is not being transparent with paying customers about deeper problems
  • The outage may be part of a broader Microsoft/Azure infrastructure failure
17.Cursor launches Origin, GitHub alternative(cursor.com)
582 points by tomasreimers 4 days ago | 439 comments | permalink
tl;dr: Cursor launched Origin, a GitHub-alternative code hosting service now in early beta for paid users, offering repos, pull requests, code browsing, and two-way GitHub sync. Repos live at cursor.com/codebase/[name], with built-in Cursor agents that can answer questions, make changes, and update PRs directly from the browsing interface. Origin also supports app extensions, with existing integrations for Vercel (preview deployments), Depot, and Buildkite (CI running GitHub Actions).
HN Discussion:
  • Decentralized or federated alternatives like Radicle, Forgejo, or Tangled would be a better path than another centralized service
  • Cursor's ownership ties to Musk/SpaceX create data privacy and supply chain risks
  • Cursor's valuation and product quality don't justify launching this; it's not a real GitHub alternative yet
  • Naming it 'Origin' creates dangerous semantic confusion with git's origin remote for LLMs
  • Developer from Origin offering to answer questions about the product
18.Memory prices climb 500% in 12 months(tomshardware.com)
657 points by haunter 4 days ago | 545 comments | permalink
tl;dr: DDR5 memory prices have surged roughly 350-500% year-over-year, with a 64GB DDR5-5600 kit jumping from under $200 to over $1,100, while DDR4 has climbed 120-180% as builders flee to older platforms. Hyperscalers have reportedly locked up nearly all 2027 DRAM production for AI datacenters, leaving consumer markets with scraps. Industry executives from SK Hynix and ADATA warn the shortage could worsen through 2027 and potentially persist for a decade.
HN Discussion:
  • Manufacturers may be underinvesting in capacity due to skepticism about sustained AI demand
  • ~The shortage could positively force developers to optimize memory usage again
  • This crisis signals the end of personal computing as we're pushed toward thin clients
  • The AI boom's effects on component prices are turning PC enthusiasts against AI
  • Personal anecdotes confirming builders are delaying purchases and sticking with older hardware
19.Claude Code May–August 2026 weekly limits promotion(support.claude.com)
292 points by tyre 3 days ago | 259 comments | permalink
tl;dr: From May 13 through August 31, 2026, Anthropic is automatically boosting weekly Claude Code usage limits by 50% for Pro, Max, Team, and legacy seat-based Enterprise plans. Free plans and consumption-based Enterprise seats are excluded, and 5-hour limits are unaffected. The increase applies across CLI, IDE extensions, desktop, and web, and limits revert to standard levels after the promotion ends.
HN Discussion:
  • Anthropic's token-heavy approach will lose to OpenAI's more efficient strategy long-term
  • Switching to Codex/GPT because Claude limits are too restrictive and quality has declined
  • ~Constant uncertainty about limit extensions is frustrating for paying subscribers
  • Clarification that limits are being made permanent, not reverting
  • Claude Code has been genuinely useful for real coding/maintenance tasks
20.Using the railway network as a flatbed scanner(philo.gay)
441 points by otherayden 3 days ago | 73 comments | permalink
tl;dr: A hobbyist mounted an industrial linear scanning camera (Basler ruL2048-19gm, ~19,000 lines/sec) out of trains and ferries to capture ultra-wide "flatbed scanner"-style images, using an accelerometer to correct for variable speed. Major challenges included integrating accelerometer data to derive velocity (GPS proved unreliable), handling parallax by tuning distance-per-pixel per segment, and fixing IR sensitivity and RGB channel misalignment on a color version. Postprocessing was rewritten in NumPy for a modular pipeline, and the resulting gigapixel TIFFs are served via OpenSeadragon tiles.
HN Discussion:
  • Sharing similar personal projects and prior experiments with line/slit scanning
  • Offering technical solutions to challenges mentioned in the article (image viewing, speed detection)
  • Appreciating the artistic and creative aspects of the project
  • Elaborating on the unique optical properties of line scan photography
  • Suggesting extensions or applications of the technique to other contexts