1. Show HN: Whispering – Open-source, local-first dictation you can trust
Total comment counts : 18
Summary
The text says feedback is read and taken seriously, directs users to documentation for available qualifiers, and notes repeated loading errors that require the page to be reloaded.
Top 1 Comment Summary
The writer is looking for a tool like this, has tried Whisprflow and Aqua Voice, wants to use their own API key and store more context locally, and asks how data is stored and how to access it.
Top 2 Comment Summary
The text advocates for open-source, local-first apps that interoperate, built on Epicenter’s model of storing data in plaintext and SQLite to form a transparent, trustworthy shared memory. The speaker enthusiastically supports this approach and plans to explore TTS with Whispering, inspired by Epicenter, and to contribute to related apps. They’ve starred the repo and applauded the authors for publishing and for YC acceptance.
2. Left to Right Programming
Total comment counts : 30
Summary
The piece argues that Python’s list comprehensions are ergonomically poor because editors can’t autocomplete or reveal types mid-type, hindering discovery. It praises Rust’s syntax for enabling immediate editor suggestions and progressive disclosure: as you type, you learn what exists and are guided toward correct constructs (the “Pit of Success”). It contrasts this with C’s lack of methods on structs, causing guesswork, and notes similar discoverability problems in Python and JavaScript. The core idea: languages should surface relevant options as you type to ease use and learning.
Top 1 Comment Summary
The piece argues that SQL’s age shows in its style, advocating starting queries with FROM to quickly determine involved entities and help editors. It suggests an order of FROM -> SELECT -> WHERE, since SELECT names columns that WHERE will reference. It even recommends avoiding SELECT * and writing FROM table with the select clause implied. The narrator ends with humorous self-deprecation.
Top 2 Comment Summary
The article notes that Python lacks a pipe operator, which the author valued when moving from Mathematica to R because it makes chained data-transformation code more readable. They wonder if a pipe would help in other Python contexts and question why Python hasn’t adopted such an operator.
3. Show HN: I built an app to block Shorts and Reels
Total comment counts : 50
Summary
The article promotes blocking short-form content across major platforms (Instagram, Facebook, Reddit, YouTube) and using an “antiscroll mode” to set scrolling limits. The aim is a distraction-free experience with no ads, reels, or shorts.
Top 1 Comment Summary
User self-hosted FreshRSS with extensions to aggregate social feeds and YouTube subscriptions. They now use a simple, chronological UI with no recommendations, algorithms, or infinite scrolling. Discovery happens elsewhere with intentional browsing, while consumption is divorced from algorithms. The change, they report, improves their consumption habits and overall mental health, feeling a material positive impact.
Top 2 Comment Summary
The article argues that apps are deliberately designed to bypass our self-control, turning simple scrolling into an addictive habit. It suggests guardrails for scrolling may feel unnecessary but are needed to curb this behavior. The author also links to a deeper dive on swipe-scroll-repeat addiction.
4. FFmpeg Assembly Language Lessons
Total comment counts : 11
Summary
FFmpeg Assembly Language Lessons is an educational series on writing assembly for FFmpeg, giving a practical grounding in how assembly works on computers. The linked Git repository holds lessons and (not-yet-uploaded) assignments, with the aim that readers can contribute to FFmpeg after completing them. A Discord server is available for questions, and translations exist. The page also notes loading errors (There was an error while loading), indicating site issues.
Top 1 Comment Summary
The author is awed by FFMPEG’s enormous scale, notes that even small improvements can save vast amounts of compute time, and regards the project as incredibly useful.
Top 2 Comment Summary
A reference to a prior Hacker News discussion dated 2025-02-22 that drew 222 comments, with a link to the related item (ID 43140614). The actual article content isn’t included in the snippet.
5. T-Mobile claimed selling location data without consent is legal–judges disagree
Total comment counts : 6
Summary
A federal appeals court (DC Circuit) rejected T-Mobile’s bid to overturn roughly $92 million in FCC penalties for selling customers’ location data. The FCC alleged T-Mobile, Sprint (now part of T-Mobile), AT&T, and Verizon shared location info without consent and failed to protect it. The DC Circuit unanimously held the FCC acted within authority, noting location data forms an intimate record of a person’s life. T-Mobile faces $80.1 million; Sprint $12.2 million; AT&T and Verizon face $57.3 million and $46.9 million, respectively, in separate appeals. The court also ruled carriers waived jury rights by seeking direct review.
Top 1 Comment Summary
Opt-out resources by carrier:
- AT&T: use the consent/ccpa/dnsatt link to opt out.
- T-Mobile: Privacy Center’s Privacy Dashboard; per account you can toggle off Profiling and automated decisions, Fraud/identity protection, sharing financial info, Analytics/reporting, Advertising options, and Do Not Sell/Share My Personal Info.
- Verizon: in My Verizon (web or app) go to Account Settings > Privacy Settings (gear icon) and disable: Custom Experience, Custom Experience Plus, Business & Marketing Insights, CPNI and Identity Verification programs.
Top 2 Comment Summary
During a cross-continental drive, the author found spam calls followed them in real time across states via their source area codes on a T-Mobile US phone. They believed they had opted out of all opt-out-able privacy settings, but after rechecking, found more than half of the settings were still enabled, lamenting default-enabled options and the need to continually monitor for new settings.
6. Anna’s Archive: An Update from the Team
Total comment counts : 37
Summary
Anna’s Archive reports intensified attacks on its mission to preserve digital texts and says it has hardened security. Since 2022, it claims to have liberated tens of millions of books, scientific articles, magazines, and newspapers, safeguarding them from disasters and budget cuts. It has scraped metadata from WorldCat, Google Books, IA-CDD, HathiTrust, and DuXiu to identify missing works. Partnerships with LibGen forks STC/Nexus and Z-Library have expanded access. A new WeLib fork is discouraged for lack of contribution. They invite volunteers and donations as hundreds of terabytes remain to be processed.
Top 1 Comment Summary
The author buys books and comics from niche sources (Anna’s Archive, readComicsOnline, #WONTTELL), becoming a top customer of three local shops. They ignore ad-driven trends, preferring to research and acquire excellent finds. Their online discoveries can force shopkeepers to locate obscure items. They wonder if they’re an exception, but argue these services help preserve freedom of choice.
Top 2 Comment Summary
The piece extols shadow libraries’ maintainers, arguing they deserve a Nobel Prize for their contributions to humanity and noting that Satoshi Nakamoto would be proud.
7. The Weight of a Cell
Total comment counts : 9
Summary
Microbes are tiny, but scientists can weigh single cells with remarkable precision. Yeast (S. cerevisiae) weigh about 100 picograms per cell, E. coli about 1 picogram. In 1953, Southern Illinois researchers used Stokes’ law—measuring how yeast sink in sugar water and filming frames—to infer mass, finding ~79 pg for yeast, close to modern estimates. Yeast are near-spherical, so Stokes’ approach works; E. coli’s rod shape complicates it. By 2010, MIT scientists devised a general tool to weigh individual bacterial cells across shapes.
Top 1 Comment Summary
Most kitchen scales are only precise to the nearest gram and aren’t very accurate overall. Scales with 0.1-gram sensitivity exist, but they’re rare and typically more expensive, especially for reliable models.
Top 2 Comment Summary
The note questions whether weighing a single cell is more accurate than counting 10^9 cells for a bulk weight, and highlights the challenge as achieving precise cell counting while excluding liquid trapped between cells.
8. Launch HN: Reality Defender (YC W22) – API for Deepfake and GenAI Detection
Total comment counts : 13
Summary
Reality Defender helps organizations combat AI-generated deepfakes. The platform offers API and SDKs (Python, TypeScript, Java, Go, Rust; open source variants) to detect deepfake and AI-manipulated media. A free tier provides 50 audio or image scans per month. Pricing starts at $399/mo with 50+ scans, plus live chat, web interface, and integrations (Zoom, Teams, Webex); custom pricing covers higher volumes, multi-seat teams, video, text, and exports. Use cases include identity verification, fraud detection, content moderation, synthetic media detection, and media analysis, plus research and testing. Insights from Reality Defender accompany the service.
Top 1 Comment Summary
The piece notes the growing concern over deepfakes and questions what happens if detection tech becomes so advanced that it’s detectable only by automation. In that future, detectors could become authoritative arbiters of truth about deepfakes. It imagines scenarios like a scandalous politician or contested video evidence in court, and asks whether deepfake-detection creators would testify as expert witnesses and how they could convince judges that a black-box output isn’t a false positive, highlighting the need for explainability and credible evidentiary standards.
Top 2 Comment Summary
The piece questions whether online AI detectors actually work, noting many tools are unreliable. It asks for a high-level discussion of evaluation criteria (what to look for), the most common failure modes, and how those issues were addressed in practice.
9. Counter-Strike: A billion-dollar game built in a dorm room
Total comment counts : 25
Summary
The article instructs readers to enable JavaScript and disable any ad blocker to access the site.
Top 1 Comment Summary
Many gamers mourn the loss of server browsers and the rise of matchmaking, which makes building small communities harder and drains serendipity. The author recalls adminning about 18 connected Counter-Strike 1.6 servers (T3Houston) that ran Warcraft 3 mods with persistent XP, an external item store, and a shared player database, plus a lively forum. He found the community by chance, formed real connections, and recalls it as unusually non-toxic. He suspects toxicity has worsened as management tools faded, signing off as Stealth Penguin.
Top 2 Comment Summary
Sorry — I can’t access paywalled content from that link. If you paste the article text (or a few key passages) or share the main points you want included, I’ll summarize it in under 100 words. Alternatively, tell me the specific aspects you want emphasized (e.g., how Minh Le shaped Counter-Strike, the article’s main conclusion), and I’ll craft a concise summary. If you’d like a general background instead, I can provide a brief overview of Minh Le and Counter-Strike’s origins.
10. Show HN: We started building an AI dev tool but it turned into a Sims-style game
Total comment counts : 19
Summary
error
Top 1 Comment Summary
The PSA warns viewers that the video does have commentary, but the audio is extremely quiet and requires turning up the volume. The author notes they initially thought the video had no audio.
Top 2 Comment Summary
The author praises embracing unpredictability rather than sanitizing it, arguing that turning tooling into something playful and emergent offers a more durable approach. They welcome the surprising agent behaviors and ask what has been the strangest or funniest observed so far.