0:00 / 2:17aideep dive Cursor users push back after dollar-cost usage data disappears#
Cursor users on the company forum report that the Usage page and CSV export stopped showing dollar costs for self-serve plans, including Teams, replacing per-request and per-model cost views with token-based reporting. Cursor staff in the thread said this was an intentional design change: Enterprise plans still show dollar amounts, while self-serve plans are now token-based in the Usage view; Teams admins are pointed to other admin spending views or APIs for supported cost data. The change matters because users say they relied on per-day, per-model, and per-request dollar figures to manage budgets and compare model efficiency, not just to reconcile final bills.
Discussion: Mixed — The discussion is skeptical of Cursor’s change and more broadly wary of AI coding-tool costs, but much of the HN thread turns into practical debate about token efficiency across agent harnesses. Several commenters question Cursor’s current value versus Claude Code, Codex, terminal tools, or plain IDE workflows, while a few defend Cursor’s IDE integration and model flexibility. (Frustration over reduced billing transparency, Token counts are hard to translate into budgets, Agent harnesses can vary widely in token usage)
0:00 / 1:56ai ByteDance’s Seedance 2.5 promises 30-second AI video takes#
ByteDance’s Seed team announced Seedance 2.5, a new video generation model focused on longer clips, reference-based generation, and more controllable editing. The headline claims are 30-second audio-video generations, multi-round extensions for multi-minute outputs, support for up to 30 image references, 10 video references, and 10 audio references, plus timestamp-level edits. ByteDance says the model is rolling out on Jimeng AI and Doubao Pro, with API access planned through BytePlus ModelArk; the post also acknowledges remaining issues with complex motion physics and multi-subject scene stability.
Discussion: Mixed — HN was broadly impressed by the apparent jump in video quality and coherence, with several commenters saying the examples looked close to usable for ads or indie filmmaking. But the enthusiasm was tempered by worries about misinformation, spam, deepfakes, lack of creative control, access confusion, and the fact that weights are not available. (High perceived quality and coherence, Potential use in ads, filmmaking, education, and synthetic training data, Concerns about misinformation, spam, and deepfakes)
0:00 / 1:37startups A cautionary tale from the startup founder meat grinder#
The source is a personal essay about an anonymous acquaintance, “Jim,” who went from coding bootcamp and a startup job to YC, repeated startup attempts, and then an apparent collapse after social-media hype, lifestyle inflation, and alleged party culture. The author frames Jim as an example of Silicon Valley’s founder pipeline: a system that can rapidly elevate charismatic risk-takers, while most do not become the celebrated outliers. It matters less as hard reporting than as a sharp, anecdotal critique of the incentives around founder status, networking, and “fake it till you make it” ambition.
Discussion: Mixed — HN largely treated the essay as a believable cautionary story about Silicon Valley founder culture, but commenters split on whether the lesson is recklessness, ambition, vanity, or survivorship bias. Many were skeptical of people who want the identity of being a founder more than the work of building a business, while a few defended risk-taking and noted that persistence and charisma can produce high-variance successes. (founder identity versus building real products, risk-taking and high-variance outcomes, charisma, status, and networking in startups)
0:00 / 3:14policydeep dive Canada signs a UN cybercrime treaty critics call a surveillance pact#
Michael Geist reports that Canada has signed the UN Convention against Cybercrime, after previously opposing the process and skipping the official signing ceremony less than a year ago. He argues the treaty is less a narrow cybercrime pact than a broad cross-border surveillance and electronic-evidence-sharing framework, with powers that can apply beyond cyber offences and potentially involve countries with repressive criminal laws. Signing does not bind Canada until ratification, but the piece says the quiet reversal, lack of public consultation, and possible connection to Canada’s lawful-access agenda deserve scrutiny.
Discussion: Negative — HN reaction was largely skeptical and worried, focusing on surveillance creep, vague “protect the children” justifications, and frustration that digital-rights law keeps moving toward more state access rather than stronger protections. A minority added context that signing is not the same as ratification and that other allies, including Australia, the EU, and the UK, have also signed. (Concern over cross-border surveillance and electronic evidence sharing, Distrust of political messaging around child protection and cybercrime, Calls for stronger digital rights, privacy, and anti-dragnet protections)
0:00 / 2:34generalfeature Why elevators feel smarter—or dumber—than they look#
John’s interactive explainer breaks down elevator scheduling, starting with single-car SCAN and LOOK algorithms and moving into multi-car dispatch systems such as Otis’s Relative System Response, which re-optimizes assignments every few seconds. The counterintuitive punchline is that more information is not always better: destination-dispatch kiosks can lose flexibility because riders are locked to a specific car, while simpler up/down calls can be rebalanced as conditions change. The piece matters because it turns a familiar annoyance into a concrete lesson in optimization, queues, UX, and the tradeoff between smart systems and adaptable ones.
Discussion: Positive — HN largely loved the explainer and its smooth interactive demos, while using the thread to trade elevator war stories and argue about real-world edge cases. The sharpest debate was over destination dispatch: some commenters agreed it can feel rigid and confusing, while others said it works well for clustered lobby traffic, tall buildings, or reducing in-car stops. (Praise for the article’s animations and clarity, Debate over destination dispatch versus up/down buttons, Real-world frustration with full cars, poor capacity handling, and disconnected elevator banks)
0:00 / 3:03generaldeep dive The old fight over Google, RSS, and the open web resurfaces#↻ from 2023
This 2023 Open RSS essay resurfaced on HN today, arguing that Google repeatedly embraced RSS in products such as Chrome, FeedBurner, Google Reader, Google Alerts, and Google News, then removed or reduced support after users had come to rely on it. The piece frames those decisions as damaging to confidence in RSS and, by extension, to the open web. It matters because RSS remains a simple, non-algorithmic way to follow publishers, blogs, and podcasts outside proprietary social feeds.
Discussion: Negative — The HN thread is overwhelmingly nostalgic and critical of Google, especially over the shutdown of Google Reader and the broader move from open websites to platform-controlled feeds. A few commenters push back with practical points that RSS is still usable through independent readers, browser extensions, and self-hosted tools, but the dominant mood is loss of trust in big-tech stewardship of open web infrastructure. (Anger over Google Reader’s shutdown, Nostalgia for the early web and personal websites, Concern about walled gardens and ad-driven platforms)
0:00 / 1:53general HN turns RamenHaus into a ramen culture debate#
RamenHaus, a standalone site at ramen.haus, reached the HN front page and sparked a mostly food-culture discussion rather than a technical postmortem. The supplied material does not include article text from the site itself, so the grounded takeaway is the community reaction: readers debated ramen as craft, fast food, and export, especially whether Western-city pricing can meet expectations set by Japan. A few comments alluded to the site's city choices and 3D presentation, but the available evidence does not establish the site's implementation details or full scope.
Discussion: Mixed — The discussion was enthusiastic about ramen as a craft and cultural object, but skeptical about the value and quality of ramen in many Western cities. Much of the thread drifted from the submitted site into personal recommendations, Japan comparisons, restaurant economics, and the film Tampopo. (ramen quality in Japan versus abroad, high restaurant prices in Western cities, Tampopo as ramen-culture touchstone)
0:00 / 2:09software Diátaxis gives technical docs a four-part map#
Diátaxis is a framework for technical documentation that organizes material around four user needs: tutorials, how-to guides, technical reference, and explanation. The site argues that separating those forms helps teams decide what to write, how to write it, and where it belongs, without imposing a specific toolchain. The HN interest is less about new software and more about a practical information architecture that many engineering teams can apply to messy docs.
Discussion: Positive — HN was broadly warm toward Diátaxis, with several commenters saying it helped them structure real documentation work and made writing clearer. The caveats were practical rather than hostile: don’t treat it as gospel, understand the framework before reorganizing docs, and note that other documentation models exist. A smaller thread connected Diátaxis to LLM-generated first drafts, where commenters found the framework useful as a promptable structure. (strong practitioner endorsement, useful but not dogma, LLM-assisted documentation drafts)