Mole for Mac 1.12 Ships; Podscan Adds Stock Ticker Tracking

Key Takeaways

1. Mole for Mac 1.12 ships with lower resource use, smoother long lists, and more reliable updates, directly built around user feedback.
2. Podscan adds stock-ticker tracking for public companies after funded institutions and funds requested it.
3. LLM text watermarking remains weak as independent verification: provider-only detection, rewrite sensitivity, and semantic distortion remain.
4. Builders flag AI UX issues: fast models should narrate long tasks, and silent model downgrades can degrade output quality.
5. Tibo's one-prompt video pipeline with Claude, Revid CLI, and Fable 5 shows AI video cost and friction falling.
6. Levelsio's Moccamaster plastic critique and switch to a plastic-free clone keep material health in product design conversations.

1. Product Launches and User-Driven Development

1. Mole for Mac 1.12 is out with a focus on "less Mac usage, more honest presentation": fewer sampling wakes, menu bar CPU no longer climbs, long lists stay smooth, cleanup now covers Dia/Zed/expired caches without stalling, App Store and Homebrew updates are more reliable, Battery Care supports on-demand charging to 100%, and the menu bar runner has been redrawn pixel-clear. The release is a direct response to real usage and performance complaints. — via 1
2. Mole's roadmap is being steered by direct user feedback: a 70-year-old British user who accidentally bought two copies refused a refund, pushing the author to make the product better and friendlier; a low-vision user reported the hard-coded dark mode made it unusable, so light mode/system-follow is planned; a US user corrected the temperature default assumption and advised Celsius-by-default with Fahrenheit optional, while suggesting the price be raised from $32 to $39 with 5 activation slots to better match CleanMyMac; a German lecturer and a Hungarian doctor pushed education discounts and country-specific pricing. The lesson the author draws: never stop talking to users because they will improve your product. — via 1
3. Podscan added a way to track company mentions by stock ticker, covering all public companies, built because financial institutions and funds specifically asked for it. For Arvid Kahl, this is the product lesson in action: when funded customers request a feature, you implement it. — via 1

2. AI, Model Behavior, and Agents

1. Arvid Kahl's deep-dive on LLM text watermarking explains that watermarking works by slightly perturbing the statistics of the next token and is designed to be detectable only by the model provider holding the key. The mechanism weakens after rewriting, cannot be independently verified, only applies to specific model outputs, and subtly distorts text semantics — in his words, it's essentially a "self-censorship + trust me" form of verification. — via 1
2. Ben Tossell predicts fast models that hide their thinking may be the future, but argues that when a task will take more than a few seconds, the model should stream text explaining what it's doing to avoid dead pauses. He also amplified "Agent Effectiveness" work that measures cycle time, work prioritization, and delivered outcomes to understand AI ROI in factory settings. — via 1 2
3. Jon Yongfook noticed Claude silently downshifting his session to Opus 4.7, which produced noticeably poor output; switching to Claude 5 restored quality. The episode is a useful warning that model routing and silent downgrades can quietly degrade results in AI-assisted work. — via 1
4. Tibo shared a single-prompt video generation workflow: hand a post to Claude, run the Revid CLI, let Fable 5 handle the back-and-forth, and keep the resulting video editable. He credits Revid with halving AI video price, switching to per-second billing, and auto-refunding failed runs. — via 1

3. Design, Regulation, and Market Signals

1. Levelsio criticized premium coffee maker Moccamaster for using an all-plastic design in 2026 and initially said no alternative existed; he later found a plastic-free clone and switched. He also flagged Nespresso's aluminum pods, citing an (unverified) association between aluminum exposure and dementia risk — a reminder that material choices are becoming a product-decision factor. — via 1 2 3
2. Jack Ellis says online SaaS doom doesn't match his real conversations: the businesses he talks to are doing well. His advice: don't trust loud internet narratives over direct market contact. — via 1
3. Levelsio also amplified criticism of EU regulation for dragging down the economy while spawning a packaging-waste compliance consulting industry. The takeaway for builders is that regulatory burden is both a headwind and a source of new services. — via 1