Hasty Briefs - Daily Tech News Brief

Bilingual

MTV Didn't Die of Bad Taste: It Died the Day the Doors Multiplied
4 hours ago
- MTV was not a spontaneous cultural movement but a business assembled by radio and cable professionals, funded by Warner Communications and American Express.
- The channel's raw material (music videos) was provided free by record labels as advertising, creating an enviable cost structure.
- MTV's real customer was cable operators, not viewers; the 'I Want My MTV' campaign pressured operators to carry the channel for a per-subscriber fee.
- The per-subscriber fee provided recurring, predictable revenue that formed the floor of MTV's business model.
- MTV gained power as the gatekeeper for music success, exemplified by the Michael Jackson 'Billie Jean' airplay negotiation.
- The shift to reality shows like 'The Real World' was a rational move to own intellectual property and generate more ad revenue than music videos.
- The rise of Napster, iTunes, YouTube, and Vevo destroyed MTV's exclusivity over music video distribution, eroding its leverage.
- MTV's decline mirrors the fate of any platform that depends on a distribution channel it doesn't control; the business model failed when doors multiplied.
- Key lessons: customer is rarely who it seems, free content is a trap, owned IP is crucial, and cultural brands die from lack of leverage, not bad taste.
Show HN: Breadcrumb, record everything on your mac + context manager for AI
5 hours ago
- - Time Machine local snapshots are consuming 197 GB; use tmutil listlocalsnapshots / to view and tmutil deletelocalsnapshots / to clear them.
- - Meeting follow-ups: Ella wants retailer bug addressed ASAP, Kai asks about loyalty dashboard (ships Friday), Luke needs sandbox login after the call.
- - Yesterday's work: 9h12m at Mac with one meeting; tasks included email/support, standup, loyalty dashboard, Breadcrumb landing page, and Slack catch-up.
- - AI chat sessions can be continued; new chats can pick up context from long sessions, with session details (durations, turns) shown.
- - User can set permanent rules for AI (e.g., never push main without explicit approval) and create one-word shortcuts for recurring reports (e.g., \fb-hours).
- - Privacy: Breadcrumb runs locally, encrypted with user key, no voiceprints, no telemetry, optional exclusions, and works offline; free and open to all.
- - Usage stats show 14h22m total, with 10h41m working, 3h41m AI unattended, long streak 1h54m, app usage (Claude 5h58m, Discord, Slack) and web time.
Show HN: Rhun, an open-source code editor written in assembly
3 hours ago
- AI handles heavy tasks, keeping the code editor lightweight and fast.
- Rhun is a small, fast code editor written in assembly.
- Features include tabs, syntax highlighting, fuzzy search, and optional Vim mode.
- Users can run builds, shell commands, and coding agents alongside their code.
- Supports reviewing diffs, staging changes, and committing with AI-assisted drafting (local models or Claude Code/Codex).
- AI is optional; users can follow project sessions from Claude Code and Codex.
- Offers 39 light and dark themes, with a preview available.
- Installation via terminal command for Linux/macOS or ZIP download for Windows.
ArXiv's Updated Rate Limit Policy
3 hours ago
- arXiv is implementing a new rate limiting policy starting October 1, 2026, to address the surge in submissions driven by advanced AI tools.
- Submissions have dramatically increased: from 9,869 in September 2016 to 40,363 in September 2026, placing a heavy burden on volunteer moderators.
- AI tools are enabling low-quality submissions such as narrow-scope, salami, and AI-written papers, which consume disproportionate moderator time.
- The new policy limits submitters to two submissions per calendar month, with a maximum of three active submissions at any time.
- Rejected submissions count toward the monthly limit to encourage careful submission practices, while deleted papers before announcement do not.
- The policy applies across all categories and only to the submitter, not co-authors, aiming to equitably distribute moderator effort and protect the arXiv corpus.
- arXiv will monitor the policy’s impact and adjust as needed, with the goal of maintaining open, fast, and free access while ensuring scholarly quality.
Car Is a Smartphone on Wheels. Here's Who's Listening
3 hours ago
- First large-scale measurement study of the connected-vehicle ecosystem was conducted by partnering with Consumer Reports to test 21 vehicles and 30 companion mobile apps.
- Vehicles and apps send private consumer data to first and third parties via Wi-Fi and cellular, often sharing or selling data without consumer knowledge or control.
- Researchers blocked cellular signals using a Faraday tent to force Wi-Fi communication, revealing data flows that would normally go over cellular.
- The study found that consumers have no control over their data once it leaves their device, highlighting the need for continued scrutiny.
- The work is peer-reviewed and will be published at IMC '26, with a team from Northeastern University.
Consumer AI Fatigue Is Overwhelming, Local Businesses Should Reconsider Use
6 hours ago
- AI is blurring the line between real and computer-generated content, causing legal issues and wrongful arrests.
- Fake AI-generated news pages on social media spread misinformation without repercussions, deceiving the public.
- Local businesses increasingly use AI for graphics, but customers prefer authentic, human-created designs.
- Graphic design professionals view AI as a tool, not a replacement, emphasizing critical thinking and communication skills.
- Consumers show strong AI fatigue, preferring unique, personalized content from local designers.
The GPU Black Market That Washington Can't Shut Down
6 hours ago
- Despite US export bans on advanced AI chips to China, a thriving black market moves billions of dollars in NVIDIA GPUs through shell companies, Southeast Asian middlemen, and fake entities.
- Documented smuggling cases include Supermicro co-founder Liaw allegedly diverting $2.5 billion, Operation Gatekeeper using a Houston shell company, and Janford Realtor LLC buying chips under a real estate cover.
- Chinese companies exploit loopholes such as remote access to overseas data centers (e.g., ByteDance renting from Oracle) and building massive data centers in Malaysia (Johor) to serve Chinese customers.
- China aggressively poaches Taiwanese chip engineers, offering triple salaries and using front companies, and has validated a domestic EUV lithography prototype as part of a state-driven 'Manhattan Project'.
- US enforcement is severely underfunded: BIS has fewer than 600 employees, 150 agents, 2006-era IT, and a budget equivalent to two F-35 fighters, with median annual penalties of only $13 million.
- NVIDIA faces a conflict of interest: it gains revenue from sales to Chinese end users via intermediaries, and its CEO called the controls 'a failure' while lobbying for relaxation.
- Global fallout includes a surge in sovereign AI investments (e.g., Saudi Arabia, UAE, Japan, India, Europe), as countries build their own compute infrastructure to reduce dependency on US export policy.
Effect 4.0
4 hours ago
- Effect 4.0 is a major rebuild, offering improved speed, lower memory usage, and smaller bundle sizes through tree-shaking.
- The ecosystem is unified under a single package with zero runtime dependencies, reducing supply chain risks and ensuring lockstep releases.
- Effect scales from a single function to distributed systems, featuring typed errors, dependency injection, resource management, structured concurrency, and observability.
- Adoption is mainstream, used by companies of all sizes, with community projects like Alchemy and Foldkit.
- Long-term support is introduced: Effect 4.x bug fixes until September 2029 or one year after 5.0, security fixes until September 2029 or two years after 5.0.
- Future plans include stabilizing the ecosystem, refining unstable modules, and expanding native platform support and higher-level primitives.
- Resources such as a migration guide and documentation are available for users.
Pi Durable
4 hours ago
- Pi 1.0 is released, and with it comes Pi Durable, an experimental package for building long-running, durable, and malleable agents that can run anywhere.
- Pi Durable is a harness—a combination of storage and machinery—that supports concurrent conversations, survives crashes via checkpoints and task resumption, and enables multiple clients to steer the same agent.
- Key features include: running on any JavaScript runtime, checkpoints for crash recovery, concurrent conversations, pluggable extensions with hooks and tools, durable application state via typed documents, dynamic code updates without restart, and multiplayer support for real-time collaboration.
- Compaction and handoff mechanisms allow infinite conversations by summarizing older messages and resetting context when needed, while retaining searchable history.
- Pi Durable is available now as experimental; examples and instructions are provided for trying it out.
iPod of 2026
5 hours ago
- Sudo is a dedicated music player focused solely on listening to music.
- Features include playback, offline and streaming, 64GB storage expandable via microSD up to 1TB, tactile wheel control, 3.5mm and Bluetooth audio, 2.73-inch LCD touchscreen, Sudo OS, dimensions 103.5×58.6×8.5 mm, 20+ hours battery life, and USB-C charging.
- Reservation requires a $50 fully refundable deposit applied to the final purchase price.
- Expected retail price is $249, shipping at the start of Q4 2027.
- Initially available in the US, Canada, UK, Australia, and New Zealand; waitlist for other countries.
- Color and shipping details can be confirmed before order enters production.
Clef: Open-source decision models, and new RL fine-tuning platform
7 hours ago
- Cloudflare releases Clef and Clef-flash, open-source decision models hosted on Workers AI, with Apache 2.0 licenses on Hugging Face.
- Clef leads the Jev Decision Index benchmark, offering faster, smarter, and fully Jev-API-compatible decision-making models.
- Decision models return typed structured outputs with probabilities for routing, escalation, or human deferral, enabling autonomous agent workflows.
- Clef features a vision encoder, 64k context window, and lower latency compared to Jev and general LLMs (e.g., 2.2s vs 4.7s for domain classification).
- Clef is built on Qwen with a non-autoregressive architecture using two-stage attention routing and schema-bound scoring, trained with label-smoothed cross-entropy, Brier loss, and RLCD.
- Cloudflare introduces a reinforcement learning (RL) fine-tuning product and service, using AI Gateway, Workers AI, Containers, and Trainer to customize Clef for specific workloads.
- Fine-tuning enables specialized use cases like Trust & Safety, support triage, and bot classification, trading general performance for domain accuracy.
- The models are enterprise-ready with no data retention or training on customer requests, and available via developer docs and live demo benchmarks.
I'm leaving Reddit because of RSS
4 hours ago
- Reddit announced it will discontinue RSS feeds on November 13, 2026, citing large-scale scraping and automated abuse, with no replacement for non-moderation feeds.
- Moderators are directed to supported alternatives like Devvit apps, including Discord Relay, for moderation alert workflows.
- The author plans to stop using Reddit on the shutdown date unless Reddit reverses course, and will shut down their product support subreddit.
- The author removed subreddit links from business support pages, prefers Mastodon as an alternative, and is considering Lemmy.
- The author criticizes Reddit's suggestion to replace RSS with Discord as absurd because Discord is a competing social network.
- The author is skeptical of the bot-scraping rationale, noting Reddit already rate-limits RSS and could tighten limits instead of eliminating the feature.
The AI Pascal's Wager
6 hours ago
- The article presents three positions on AI-generated code in open source projects: full acceptance, a middle-ground (like Debian), and strong rejection.
- The middle-ground is argued to be unsustainable, effectively leading to full acceptance over time due to gradual infiltration of AI contributions.
- Adopting AI may increase contributions but alienates ethical contributors and creates dependence on proprietary tools; rejecting AI preserves independence but may alienate some contributors.
- Long-term risks of AI slop are compared to asbestos—difficult to remove and potentially forcing project abandonment.
- The author advocates a Pascal's wager: reject AI now to avoid future regret; if AI proves beneficial, policy can be changed later at no cost.
Books that made me a 10x engineer
5 hours ago
- The article highlights three books that shift an engineer's mindset from coding to thinking.
- The Pragmatic Programmer offers timeless wisdom: fix root causes, master tools, and don't leave broken windows.
- Thinking, Fast and Slow improves decision-making by explaining the two systems of thought and reducing cognitive biases.
- Designing Data-Intensive Applications teaches scalable system design and understanding tradeoffs like consistency vs. availability.
- These books don't focus on syntax but change perspective, helping engineers think in systems and ask better questions.
- The author recommends reading them slowly and letting the ideas reshape how you view the craft.
SlutCon
6 hours ago
- A 26-year-old lawyer from New Zealand, using pseudonym Overlocked, attends SlutCon, a three-day retreat in Berkeley aimed at improving men's seduction skills, as a compensated 'flirt girl'.
- The event is hosted by Lighthaven, a venue operated by Lightcone Infrastructure, closely linked to the 'rationalism' subculture, which emphasizes cognitive bias avoidance and Bayesian reasoning, particularly active in Silicon Valley.
- SlutCon's philosophy is to help men better seduce women through candid feedback and practice, with participants wearing wristbands for approachability and flirt girls offering gentle corrections to avoid 'crushing their spirit'.
- Tickets cost between $3,300 and $18,000, with higher tiers offering workshops and personalized feedback; attendees are predominantly white, awkward, and rationalist-adjacent, with a mix of flirt girls from various backgrounds.
- Panels focus on making attraction more calculable, such as 'Get Sexier' discussing physical traits like canthal tilt and BMI, and 'Rizz Under Uncertainty' framing rejection as data rather than personal failure.
- The author feels safe but notes instances of entitlement, like a comment about shooters and defensive reactions to feedback; she appreciates the direct communication style, contrasting it with her usual guardedness as a lawyer.
- She discovers a neurodivergent theme, with many attendees and herself being autistic, leading to a sense of community and shared experiences of alienation, yet she remains emotionally distant throughout the weekend.
- The author critiques the event's attempt to sanitize attraction through measurement, arguing that real connection requires vulnerability and inefficiency, as seen in the tenderness among flirt girls behind the scenes.
- She reflects on her own guardedness, realizing she avoided being 'solved' due to fear, finding irony in the event's goal of connection while she protected herself.
- Comments applaud the article's insight, drawing parallels to autism training for children, suggesting nervous system calming could be more effective than behavior training, and noting Aella's in-person humility.
Breaking the Cipher of the Napoleonic war briefing, unread for 217 years
6 hours ago
- A letter to General Marmont, encrypted and unread for 217 years, was finally deciphered using AI tools like GPT-6 Astra.
- The cipher was a homophonic substitution code with 155 distinct signs; only 33 letter values were previously known.
- The solution process combined transcription, simulated annealing, and cross-referencing with Napoleonic texts, producing a full reading from a single scanned plate.
- The decrypted letter, dated to March 1809, listed French and allied army positions and was a forecast of the upcoming war with Austria.
- The letter's predictions matched historical events (e.g., Austrian troop positions and the Russian response), showing the accuracy of Napoleon's intelligence.
- This breakthrough demonstrates how modern AI can rapidly solve historical ciphers that were previously considered unsolvable by human specialists.
Big Tech's Capex Is Half of Wall Street's Profit Growth
4 hours ago
- Goldman Sachs forecasts $1.2 trillion in hyperscaler capital expenditure for 2026, a 50% increase from 2025, with $1.4 trillion projected by 2028.
- Nearly half of S&P 500 earnings per share growth in 2025 came from AI-related spending, and a $250 billion deviation in capex could shift earnings growth by 6 percentage points.
- The earnings boost from AI investment is tied to the growth rate of spending; as capex growth slows from 94% to 50%, depreciation expenses will increasingly offset profits, potentially turning AI investment into a net drag by 2028.
- Big Tech booked roughly $150 billion in unrealized gains from private-company equity stakes in Q2 2025, temporarily inflating S&P 500 earnings by 12%.
- Hyperscalers need approximately $300 billion in annual AI revenue to break even on their investments, but current cloud revenue above the pre-AI trend is only about $70 billion annually.
- For hyperscalers to achieve adequate returns, AI users would need to spend roughly $1 trillion per year on AI applications—about two-thirds of global software spending.
- Goldman’s base case is a slowdown in earnings growth, not a collapse, with S&P 500 EPS expected to grow ~11% annually in 2027 and 2028, supported by AI productivity gains.
Bez: Generating a browser engine from specs and tests
5 hours ago
- Bez generates a web rendering engine from specifications and tests, avoiding manual implementation by hundreds of engineers.
- Goals include complete or tree-shaken builds, fast performance on any device, easy embedding, and fastest updates via regeneration.
- The engine works by modeling specs into candidates, running them, and checking against Chrome, Firefox, and WebKit for consistency.
- Current status shows 93% unreached functionality; CSS leads with 2.5% generated, while HTML, JavaScript, and others remain at 0%.
- Findings reveal that the three browsers mostly agree, Firefox is an outlier in certain cases, and conformance suites like WPT are usable oracles.
- Logic programming (Datalog) accurately replicated margin collapsing across 1195 offsets and caught bugs missed by geometry checks.
- About 55–60% of engine-relevant features have usable automated oracles and generatable spec prose, while 8–18% lack both.
- Open questions remain about cost, feature scope, WPT reachability without JavaScript, and boundaries of IDL-generated code.
cp: -r or -R?
a day ago
- In modern GNU coreutils, there is no difference between the -r and -R options for cp; both are identical and perform recursive directory copying.
- Historically, in the first GNU coreutils commit (1992), -r and -R differed: -r set a flag to copy regular files as regular, while -R did not.
- Other systems like OpenBSD and NetBSD still distinguish between the two, with -R preferred and -r discouraged, though -r remains different in behavior.
- POSIX.1-2024 no longer specifies the -r option for cp, recommending -R instead for consistency across utilities that perform recursive directory descent.

Hasty Briefsbeta

MTV Didn't Die of Bad Taste: It Died the Day the Doors Multiplied

Show HN: Breadcrumb, record everything on your mac + context manager for AI

Show HN: Rhun, an open-source code editor written in assembly

ArXiv's Updated Rate Limit Policy

Car Is a Smartphone on Wheels. Here's Who's Listening

Consumer AI Fatigue Is Overwhelming, Local Businesses Should Reconsider Use

The GPU Black Market That Washington Can't Shut Down

Effect 4.0

Pi Durable

iPod of 2026

Clef: Open-source decision models, and new RL fine-tuning platform

I'm leaving Reddit because of RSS

The AI Pascal's Wager

Books that made me a 10x engineer

SlutCon

Breaking the Cipher of the Napoleonic war briefing, unread for 217 years

Big Tech's Capex Is Half of Wall Street's Profit Growth

Bez: Generating a browser engine from specs and tests

cp: -r or -R?