HackerNews Digest

January 10, 2026

“Erdos problem #728 was solved more or less autonomously by AI”

The page is a Mastodon post on the Mathstodon instance titled “Terence Tao: ‘Recently, the application of AI tools to Erdos pr…’”. The content consists primarily of the Mastodon interface prompt urging users to enable JavaScript to use the web application, with a note that alternative Mastodon clients are available. An “Images and Visual Content” section is present, listing a single image placeholder (Alt text: “Mastodon”). No substantive discussion, excerpt, or commentary from Terence Tao regarding AI tools applied to Erdős problems is included in the scraped text. The page therefore contains only the title, generic Mastodon usage instructions, and a reference to an image, without further technical or contextual information.
Read full article →
The discussion reflects strong enthusiasm for AI‑assisted theorem proving, highlighting recent successes, the potential for rapid, superhuman proof generation, and the promise of accelerating research on hard problems. Simultaneously, many point out that current achievements rely heavily on expert human guidance, questioning claims of full autonomy and emphasizing the need for careful validation. Concerns are raised about over‑hyping, the role of training data, and how AI‑coauthored work will be evaluated academically, while acknowledging that progress remains incremental and field‑specific.
Read all comments →

JavaScript Demos in 140 Characters

Dwitter is an online platform that showcases JavaScript demos constrained to 140 characters of source code. The site encourages developers to create compact visual or interactive sketches, emphasizing brevity, creativity, and efficient coding techniques. Users submit their snippets, which are rendered in-browser, and can view, fork, or discuss others’ work. The 140‑character limit serves as a technical challenge, driving innovative use of language features, concise algorithms, and minimalistic graphics. Community interaction revolves around sharing these micro‑demos, exploring optimization strategies, and discovering novel visual effects achievable within the strict character budget. Dwitter thus functions as a showcase for succinct JavaScript programming and a collaborative space for experimentation with terse code.
Read full article →
The comments collectively view the 140‑character JavaScript demo platform as an inventive and enjoyable space that showcases impressive visual tricks and revives code‑golf culture. Many praise the creativity enabled by tight size limits, while noting that widespread reliance on eval‑unescape tricks feels like cheating and wishing for richer shorthand operators and safer scripting controls. Nostalgic references highlight past byte‑squeezing communities, and a few technical complaints mention site responsiveness problems on mobile browsers. Overall sentiment is positive, tempered by suggestions for clearer rules and improved usability.
Read all comments →

RTX 5090 and Raspberry Pi: Can it game?

An RTX 5090 can be connected to a Raspberry Pi 5 via an OCuLink eGPU dock, but gaming performance is limited by the Pi’s ARM CPU and its Gen 2 ×1 PCIe bandwidth (~500 MB/s). Linux drivers are patched to run under Ubuntu/Fedora, and games require multiple compatibility layers (FEX for x86 emulation, Wine/Proton, DXVK or WineD3D). Benchmarks show modern titles (e.g., Cyberpunk 2077) stall at ~15 FPS even at 1080p/Ultra‑Raytrace, while the Beelink N150 (Intel N150, 4 cores, PCIe Gen 3 ×4) reaches >50 FPS on the same settings. Older games perform better: Portal 2 runs at 4K > 60 FPS natively, Just Cause 2 demo hits ~40 FPS, and Doom: The Dark Ages is GPU‑bound at ~90 FPS but limited to ~30 FPS by the CPU. Power draw: Pi 5 stays <9 W under load; Beelink ~30 W. The conclusion is that eGPU gaming on a Pi 5 is technically possible but impractical for modern games; a low‑cost x86 SBC (Beelink) offers far better performance, while future ARM‑optimized drivers may improve viability.
Read full article →
Comments express surprise that the setup runs modern software at all, noting the novelty of attaching a high‑end GPU to a Raspberry Pi and recognizing some niche usefulness for GPU‑heavy, bandwidth‑limited workloads. However, most view the configuration as impractical for typical desktop or gaming use due to the Pi’s cost, limited CPU performance, and better value in x86 mini‑PCs or used thin‑clients. Linux/Proton compatibility is praised, while concerns about anti‑cheat support, power efficiency, and overall relevance of gaming benchmarks dominate the discussion.
Read all comments →

Flock Hardcoded the Password for America's Surveillance Infrastructure 53 Times

A default organization‑wide ArcGIS API key was hard‑coded into 53 publicly accessible JavaScript bundles served by Flock Safety. The key—created automatically when the company signed up for an ArcGIS developer account—had no referrer, IP or scope restrictions and was granted 50 “portal:app:access:item” privileges, allowing read access to private ArcGIS items. Querying the API revealed that these items host the unified “FlockOS” mapping layer, which aggregates data from roughly 12 000 law‑enforcement, community and private deployments, including: - License‑plate and people detections, hotlist alerts, and vehicle‑search records - Real‑time locations of patrol cars, body‑camera units, drones, and Raven audio sensors - Camera inventories, serial numbers, uptime, and registrant contact information - Flock911 incident layers with call IDs, transcript tokens and audio playback data A separate, unauthenticated token‑minting flaw enabled generation of valid ArcGIS tokens for the same production environment. The vulnerabilities were disclosed in November 2025; after 55 days the API key remained unrotated and the token issue unfixed. The exposure provides a single point of entry to nationwide surveillance data, raising significant intelligence‑gathering and domestic‑privacy risks.
Read full article →
The comments express strong criticism of Flock’s privacy practices and distrust of the company’s competence, portraying its surveillance technology as invasive and poorly evaluated by municipal officials. Many contributors highlight difficulties in removing installed cameras and question the legitimacy of the company’s security claims, while also noting recent efforts to improve internal security staffing. Skepticism appears regarding technical explanations in related articles, and there is a broader concern that public camera feeds should remain openly accessible, reflecting a desire for greater oversight and transparency.
Read all comments →

Show HN: Rocket Launch and Orbit Simulator

The page presents an interface for an orbital rocket simulation. It lists telemetry parameters—altitude, downrange distance, velocity, vertical and horizontal components, acceleration, max dynamic pressure (Q), drag coefficient, Mach number, mass, propellant percentage, thrust, pitch angle, apoapsis, and periapsis—each currently showing zero or default values. A “Pitch Program” section specifies a target and current pitch of 90°. Additional sections for “Manual Pitch Control” and “Orbital Burns” are present but contain no data or controls. The layout functions as a dashboard for monitoring and adjusting a rocket’s flight profile, though no active flight data or user inputs are displayed at this time.
Read full article →
The comments are overwhelmingly positive, praising the simulation’s visual appeal, educational value, and potential audience, while also noting curiosity about the guidance algorithms, AI involvement, and mathematical foundations. Users express interest in more accurate, continuous‑burn guidance and suggest enhancements such as mobile‑responsive design, refined theming, less sensitive zoom, and clearer UI cues. Several remarks question the realism of the launch trajectory and request additional features like varied orbital calculators, but overall the feedback is supportive and anticipates future iterations.
Read all comments →

How will the miracle happen today?

- The author recounts relying on spontaneous generosity while hitchhiking in New Jersey and traveling across Asia for eight years, citing numerous specific examples of strangers offering food, shelter, and hospitality. - He formulates a theory that receiving kindness requires an “open” state of need, describing generosity as an exchange where the recipient expresses humility, gratitude, and trust. - While cycling coast‑to‑coast in the United States, he systematically asked homeowners for a place to camp; none refused, and the encounters became reciprocal storytelling opportunities. - The essay distinguishes “kindness of strangers” from familiar altruism, arguing that such acts appear random yet are pervasive across cultures and not confined to any race or creed. - The author introduces the term “kindee” for those who accept gifts and suggests that deliberate willingness to be helped summons generosity, framing this as a spiritual or “pronoia” mindset—believing the universe conspires to assist. - Ultimately, he views life itself as an unearned, constant gift, emphasizing gratitude over hope as the core of genuine faith.
Read full article →
Comments display a mixed view of the essay’s themes. Many praise the nostalgic idea of reciprocal kindness in hitchhiking and emphasize the value of gratitude, openness, and spontaneous generosity as enriching human connection. Others criticize the author for exploiting privileged status, questioning the lack of tangible repayment and highlighting power imbalances, safety concerns, and the potential for naive generosity to be abused. Several reflections compare past communal practices to today’s content‑driven attention economy, noting both a loss of personal storytelling and a continuing belief in a fundamentally generous humanity.
Read all comments →

How Markdown took over the world

Markdown, created by John Gruber in 2004 with early beta testing by Aaron Swartz, was designed to simplify web formatting by letting users write plain‑text markup (e.g., `# Header`, `[link](url)`) that could be easily converted to HTML. It addressed the cumbersome HTML syntax required for blogging platforms like Movable Type, enabling faster, more reliable content creation. The format’s success stemmed from several factors: a memorable name, solving a real pain point, building on familiar email‑style conventions, and releasing during the rise of blogs, RSS, and early social media. An open‑source community quickly produced implementations and “flavors” (e.g., CommonMark, GitHub‑Flavored Markdown), while the lack of intellectual‑property restrictions encouraged widespread adoption. Today Markdown is supported in major editors (Google Docs, Microsoft Notepad, Apple Notes), messaging apps (Slack, Discord, WhatsApp), and is the de facto documentation language on platforms like GitHub, resulting in billions of Markdown files across devices and cloud services.
Read full article →
The overall view is that Markdown is widely appreciated for being plain‑text, easy to learn, portable across editors, and well suited to version control and LLM processing, making it a practical alternative to more heavyweight formats. Frequent criticism highlights the fragmented dialects, absence of a single authoritative standard, limited support for complex layouts or precise typography, and the fact that browsers do not render Markdown natively. Comparisons to reStructuredText, Org mode, and newer systems such as Typst note both advantages and shortcomings, while calls for better standardization and broader tool support are common.
Read all comments →

Scientists discover oldest poison, on 60k-year-old arrows

None
Read full article →
The comments blend technical clarification with broad historical reflection, distinguishing venom’s inertness when ingested from poison’s systemic danger and citing specific alkaloids from Boophone species. They question how archaeologists identify arrowheads versus atlatl darts, reference early bow evidence, and contextualize prehistoric weaponry within the Middle Paleolithic and Late Pleistocene. The discussion expands to modern toxic threats, noting that deliberate poisons remain prevalent, while also pondering humanity’s scale across millennia. Overall, the tone is analytical, inquisitive, and mildly critical of persistent human reliance on harmful substances.
Read all comments →

Start your meetings at 5 minutes past

An engineering manager at Google reports that his teams schedule all meetings to begin five minutes after the hour (or half‑hour). The shift addresses the common problem of meetings overrunning their allotted slots, especially in back‑to‑back sequences. By starting later, a brief buffer is created, leveraging social pressure that discourages extending past the official end time and giving participants a short transition period. The manager observes that attendees arrive promptly at the new start time, remain settled, and overall punctuality improves. Although five minutes are technically unused per meeting, the practice eliminates the typical early‑start lag and reduces cumulative delays. Adoption spread organically across the organization without a mandate, suggesting the habit’s effectiveness as a low‑overhead scheduling adjustment.
Read full article →
Comments converge on the importance of clear meeting discipline, with many emphasizing leadership’s role in setting expectations for punctual starts and endings. Experiences vary: some organizations adopt five‑minute buffers or staggered start times to accommodate overruns, while others view such “gaming” of schedules as unnecessary and prefer strict on‑the‑dot timing. Several users note frequent overruns, back‑to‑back meetings, and the need for brief breaks, whereas others argue that well‑run meetings naturally stay within allotted time and that buffer policies add complexity. Overall, consensus leans toward consistent cultural norms over ad‑hoc timing tricks.
Read all comments →

Show HN: Scroll Wikipedia like TikTok

The page is headed with the title “QuickQuack.” A visual separator (a line of equal signs) follows the title. Below the separator are four isolated entries presented in plain text, each on its own line: “Following,” “Slop,” “Ducks,” and “Storytime.” No additional sentences, descriptions, or contextual information accompany these words. The layout consists solely of the title, separator, and the four single-word items, with no hyperlinks, images, or metadata evident in the provided excerpt.
Read full article →
The idea of a TikTok‑style, AI‑generated encyclopedia receives strong enthusiasm, with many noting the appealing UI, animations, audio integration, and creative potential. Users request practical features such as tag‑based filtering, better viewport handling on small devices, accurate caption timing, and adjustable playback speed. Concerns are raised about high resource costs, rate‑limit restrictions, possible purpose dilution, and security risks from prompt injection. Overall sentiment is cautiously optimistic, acknowledging both the novelty’s promise and the need for refinements.
Read all comments →