最新

2

I formalized the Single Source of Truth (SSOT) principle in Lean 4 (~2.1k LOC, zero sorry) and proved two core results:<p>Structural SSOT is achievable only when a language provides definition-time hooks and runtime introspection. Macros&#x2F;codegen (before definition) and reflection (after definition) are insufficient. These requirements are derived, not chosen: because structural facts are fixed at definition, derivation must occur at definition time and be introspectable to verify DOF = 1.<p>Would appreciate review, critique, or independent checking of the Lean scripts.

Runs local browser instances of Meta&#x27;s SAM Audio playground so you can isolate vocals&#x2F;drums from audio of any length without running SAM locally or hosting inference.<p>- Audio &gt;29s is chunked with ffmpeg<p>- Audio chunks and prompts are submitted in parallel to the playground via Playwright<p>- Web UI for storing tracks and re-editing previous outputs<p>Demo video: <a href="https:&#x2F;&#x2F;github.com&#x2F;user-attachments&#x2F;assets&#x2F;d5d3b53d-6ac9-40fc-9776-1afc7efbe4f4" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;user-attachments&#x2F;assets&#x2F;d5d3b53d-6ac9-40f...</a>

I&#x27;m a final-year student from Malaysia.<p>Been spending some time reading up on Fraud in Fintech.<p>Would love feedback on what I&#x27;ve learned, missing gaps.

I use Plausible as a main analytics platform, but apparently I don’t have the discipline to check it regularly for all my side-projects. So it just sits there and I&#x27;m actually paying money for that just to exists.<p>To fix it for myself I decided that I needed a quick peek at current stats - something that’s always one gesture away and extremely easy so I could do this without any friction.<p>Unfortunately I use Windows, and Windows has this thing called Widgets. Apparently, that’s exactly the surface I needed. Their docs are terrible and 100% of existing widgets are useless, but the panel is fast enough for my task.<p>So I crafted Wizible - a Windows 11 widget for Plausible Analytics.<p>It’s intentionally super simple: open widgets → glance at stats → if something changes, then I open the full dashboard.<p>It&#x27;s free, use like 20MB of RAM, and should be relatively secure and pretty fast because I put some real software engineer work there and not vibe-code slop.<p>It&#x27;s available in Microsoft Store, I&#x27;m also might open-source it if people are interested.<p>Curious: what’s the ONE metric you’d want on the widget by default? (active visitors &#x2F; top pages &#x2F; referrers &#x2F; conversions?) I choose metrics based on my understanding, but maybe my understanding is wrong.<p>And: would you prefer &quot;real-time only&quot; or &quot;today vs yesterday&quot; trend at a glance? Thinking how a currently missing &quot;small size&quot; might look.

Hi HN<p>I built a small CLI to safely clean up stale GitLab branches.<p>It’s designed to be: - Safe by default (dry-run enabled) - CI-friendly (proper exit codes) - Automation-ready (JSON output) - Boring and predictable (by design)<p>I built it because repos tend to accumulate abandoned branches, and cleanup often happens too late or too manually.<p>Would love feedback, edge cases, or ideas for repo hygiene automation.

I built a dashboard to organize the unsealed indictments, court schedules, and filings regarding the United States v. Nicolás Maduro case in the Southern District of New York.

Hi HN,<p>We are working on a document vault designed for people and organizations who cannot accept cloud exposure.<p>The system is intentionally boring in some ways: • No required accounts • No cloud dependency for core functionality • Fully offline operation • Local encryption • Air-gapped storage • Encrypted export and controlled, encrypted printing<p>The printing piece is why we started this. In many environments, printing is still unavoidable, and it remains one of the largest data-leak vectors. Most privacy tools stop at storage and ignore output entirely.<p>This is not meant to replace cloud storage for everyone. It is for cases where the threat model assumes: • Networks are hostile • Cloud accounts will eventually be compromised • Convenience must sometimes be traded for control<p>We are explicitly not claiming: • “Unhackable” • “Military-grade” • “Zero risk”<p>We are trying to minimize attack surface and failure modes, not eliminate them.<p>We would genuinely value feedback on: • Threat model blind spots • Encrypted printing assumptions • Physical access risks • Update and key management strategies • What would make you immediately distrust this<p>If this sounds like something you would never use, that is also useful feedback.<p>Thanks.