<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>StillWorks</title>
  <link>https://stillworks.dev/</link>
  <description>New launches worth your time — picked every morning from ~27 boards, code hosts and crowdfunding sites.</description>
  <language>en</language>
  <atom:link href="https://stillworks.dev/feed.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>VPNonly — Routes only chosen Mac apps through a VPN, leaving the rest of the Mac untouched</title>
    <link>https://stillworks.dev/products/p/vpnonly/</link>
    <guid>https://stillworks.dev/products/p/vpnonly/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>VPNonly lets you put individual Mac apps on a VPN while everything else stays on your normal connection, using WireGuard and the built-in macOS firewall. It works with NordVPN, Mullvad, Proton, IVPN, AirVPN or any WireGuard config, but Safari can&#x27;t be routed (WebKit hands connections to system processes) and it&#x27;s Apple Silicon only. It&#x27;s $19 once, or free from the terminal, versus rivals like SplitTunnel.app that put your whole Mac on the VPN first and charge $49 a year.</description>
  </item>
  <item>
    <title>Aside — AI browser that logs into your accounts and finishes tasks other agents give up on</title>
    <link>https://stillworks.dev/products/p/aside/</link>
    <guid>https://stillworks.dev/products/p/aside/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Aside is a browser agent that works across your logged-in websites — email, dashboards, internal tools — to do multi-step work like signing in, replying to messages, or prepping notes from your browsing history. Credentials autofill through its Vault without the AI seeing them, and sensitive actions like payments or posts wait for your manual confirmation, with every credential use logged. It claims #1 on three browser agent benchmarks (Online-Mind2Web, BU-Bench-V1, Odyssey), ahead of Browser Use, GPT-5.4, Claude Opus 4.8 and ChatGPT Atlas.</description>
  </item>
  <item>
    <title>miinideck — Turns an HTML file, ZIP, PDF or image into a private, unsearchable share link</title>
    <link>https://stillworks.dev/products/p/miinideck/</link>
    <guid>https://stillworks.dev/products/p/miinideck/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>miinideck hosts HTML, ZIP, PDF, PNG or JPEG files (up to 50 MB) and gives you a link in about 10 seconds, with no signup required to try it. Every link is excluded from search engines by default, and you can add a password or set an expiry from 1 day to a custom date. Unlike generic HTML hosts that publish to the open web, or an AI tool&#x27;s own share button that&#x27;s often public or login-gated, miinideck defaults to private delivery.</description>
  </item>
  <item>
    <title>Obra — Turns job-site receipts from WhatsApp, email or photos into live project profit numbers</title>
    <link>https://stillworks.dev/products/p/obra/</link>
    <guid>https://stillworks.dev/products/p/obra/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Obra captures receipts from WhatsApp, email or photo upload, has AI read amount, vendor, date, category and tax, and waits for a human to confirm before anything posts to the ledger. It tracks cash collected versus recorded cost per project and per client, keeps totals bucketed by currency instead of mixing them, and flags budget burn at 80% and 100% thresholds. You can also ask Claude or ChatGPT how a specific job is doing, since the same numbers feed the AI assistant.</description>
  </item>
  <item>
    <title>Munen — Screen-time app that adjusts friction based on how you actually used your phone yesterday</title>
    <link>https://stillworks.dev/products/p/munen/</link>
    <guid>https://stillworks.dev/products/p/munen/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Munen doesn&#x27;t block apps outright — it computes a daily &#x27;zone&#x27; from screen time, nighttime use and app opens, then adds a pause of varying length before you open a distraction, always leaving the choice to you. It cites specific peer-reviewed studies behind its friction approach (like a CHI 2024 paper showing subtle friction cut usage 15.6% more than a timed lockout) while explicitly saying it doesn&#x27;t promise you&#x27;ll get those same results. Journal entries and metrics stay on your phone, and only text sent to an AI feature you turn on leaves the device.</description>
  </item>
  <item>
    <title>Baby Monitor Timmy — Turns two phones into an encrypted baby monitor with no account or cloud recording</title>
    <link>https://stillworks.dev/products/p/baby-monitor-timmy/</link>
    <guid>https://stillworks.dev/products/p/baby-monitor-timmy/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Baby Monitor Timmy connects two Android or iOS phones over Wi-Fi or mobile data, encrypting audio and optional video with no cloud recordings, tracking or ads. Setup requires confirming a matching safety number on both phones before you assign baby and parent roles, and you can save up to eight paired connections for later. Without buying Premium, live monitoring is capped at a total of 10 minutes before free transmission ends, and the app states outright it&#x27;s a technical aid, not a substitute for personal supervision.</description>
  </item>
  <item>
    <title>Lidar — Scans Reddit daily for people describing your product&#x27;s problem and sends you the threads</title>
    <link>https://stillworks.dev/products/p/lidar/</link>
    <guid>https://stillworks.dev/products/p/lidar/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Lidar reads your site, works out who it&#x27;s for, picks relevant subreddits, and scores new posts there against your product daily, surfacing only the ones describing your specific problem with a quote and a link to reply. You write and send the reply yourself — Lidar never posts or messages on your behalf, and it never connects a Reddit account. It&#x27;s €14 a month per product after a 7-day free trial with no card required upfront, and the maker calls it explicitly a &#x27;high-intent, low-volume&#x27; channel — a handful of matches a day for most products.</description>
  </item>
  <item>
    <title>One Guy Consulting — Runs a 12-step HIPAA compliance process with a person behind it, not just software</title>
    <link>https://stillworks.dev/products/p/one-guy-consulting/</link>
    <guid>https://stillworks.dev/products/p/one-guy-consulting/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>One Guy Consulting walks an organization through a 12-step HIPAA process — risk assessment, auto-generated gap analysis and remediation plans, employee training invites, AI-drafted policies, vendor management, and annual renewal reminders — automating the repetitive matching of gaps to regulations. Unlike compliance platforms that sell software and leave interpretation to you, this pairs the tool with a person&#x27;s guidance, and pitches two pricing plans instead of a five-figure invoice. Renewals run on an annual cadence, and it commits to reminding organizations when items come due again.</description>
  </item>
  <item>
    <title>QRBliss — Generates logo-branded QR codes with contrast checked automatically, no signup</title>
    <link>https://stillworks.dev/products/p/qrbliss/</link>
    <guid>https://stillworks.dev/products/p/qrbliss/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>QRBliss extracts your logo&#x27;s color palette client-side, verifies contrast stays at or above 3:1 for scan reliability, and exports a branded QR code as PNG, SVG or print-ready PDF in under 2 seconds, with no signup or watermark required to try it. It gives 15 active dynamic codes free forever, with editable destinations and basic analytics, versus competitors it says charge $15/month for one. It also states it drops zero tracking cookies on a fresh visit and uses hashed IPs with only derived signals.</description>
  </item>
  <item>
    <title>BrowserGateway — Routes Playwright, Puppeteer and MCP browser sessions across multiple cloud browser providers</title>
    <link>https://stillworks.dev/products/p/browsergateway/</link>
    <guid>https://stillworks.dev/products/p/browsergateway/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>BrowserGateway is a router that fans requests from Puppeteer, Playwright, Stagehand, browser-use or any MCP client out across Browserless, Steel, Browserbase, Lightpanda or your own Chrome, with automatic failover and session replay behind one endpoint. It&#x27;s a drop-in swap — change the connection URL and keep the rest of your script — and self-hosting the gateway and browserserve runtime is MIT-licensed and free, while the managed cloud gives free BYOK routing up to 10,000 sessions a month. Session profiles (cookies, localStorage, IndexedDB) are captured and encrypted with AES-256-GCM at rest, so a login can be replayed on a different provider than where it was captured.</description>
  </item>
  <item>
    <title>Giraffe — Keeps the decisions and current facts from your ChatGPT and Claude chats organized by project</title>
    <link>https://stillworks.dev/products/p/giraffe/</link>
    <guid>https://stillworks.dev/products/p/giraffe/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Giraffe sorts what you tell it to save from an AI chat — current values, decisions, open questions — into per-project records instead of leaving them buried in transcripts, and keeps a history when a value changes (like a contract renewing at $48,000/year down from $56,000). Nothing is saved unless you explicitly ask, and every project exports as Markdown and JSON. It&#x27;s built for the moment you&#x27;d otherwise re-explain context to a different assistant, like reading an Acme project into Gemini before drafting a renewal email.</description>
  </item>
  <item>
    <title>Bangermeter | Tweet Scorer — Scores X posts 0-100 using X&#x27;s own published ranking weights, entirely in your browser</title>
    <link>https://stillworks.dev/products/p/bangermeter-tweet-scorer/</link>
    <guid>https://stillworks.dev/products/p/bangermeter-tweet-scorer/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Bangermeter is a free Chrome extension that badges every post in your X feed with a 0-100 score, using the real production ranking weights X published on August 13, 2026, each cited to its line in X&#x27;s own param.rs source. It makes zero network requests and collects no data, ships MIT open source, and backs its weight table with 112 passing self-tests asserting each coefficient against the source line by line. It also scores your own draft as you write it, flagging things like three-plus hashtags or &#x27;like if&#x27; bait that the algorithm penalizes.</description>
  </item>
  <item>
    <title>feder-cr/AIHawk — Open-source browser agent that automates web tasks in plain English without triggering anti-bot checks</title>
    <link>https://stillworks.dev/products/p/feder-cr-aihawk/</link>
    <guid>https://stillworks.dev/products/p/feder-cr-aihawk/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>AIHawk drives a stealth Firefox browser to complete plain-language tasks — like checking flight prices across a date range — either as an MCP server for coding agents (Claude Code, Codex, Gemini CLI) or as a standalone web UI backed by your own OpenRouter key. It moves the mouse and types like a person rather than calling APIs, supports pinning a specific engine binary and browser identity via a seed for reproducibility, and reads config from a local .env that never overrides a flag you already set. The docs are explicit that binding the local server to anything beyond 127.0.0.1 exposes an interface with no authentication.</description>
  </item>
  <item>
    <title>spotifast — Lightweight native Rust client for Spotify with local playback and Connect control</title>
    <link>https://stillworks.dev/products/p/spotifast/</link>
    <guid>https://stillworks.dev/products/p/spotifast/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>Spotifast is a native Rust app (formerly Fastpotify) for Linux, macOS and Windows that starts in under a second and uses 100-250MB of RAM, playing locally at up to 320kbps or controlling Spotify Connect devices from the same window. It also doubles as a Winamp-style mini player that loads classic .wsz skins, and includes a MilkDrop visualizer with over 10,000 presets downloaded on first use. It&#x27;s MIT-licensed and open source, with docs explaining exactly what it connects to and what credentials it stores.</description>
  </item>
  <item>
    <title>Oregon Home Inspection Guide — Free reference on Oregon home inspection law, every fact dated and cited to the source statute</title>
    <link>https://stillworks.dev/products/p/oregon-home-inspection-guide/</link>
    <guid>https://stillworks.dev/products/p/oregon-home-inspection-guide/</guid>
    <pubDate>Tue, 15 Sep 2026 02:22:23 +0000</pubDate>
    <description>This guide tracks Oregon home inspector certification, licensing fees, exam details and the 2025 wildfire rule changes (SB 83), citing each fact to its statute or agency source with a verification date, like ORS 701.445(1) for the two-step certification requirement. It calls out where official documents contain drafting errors, like an OAR cross-reference to the wrong subsection, and separately tracks which of Oregon&#x27;s jurisdictions have actually adopted the R327 wildfire code versus jurisdictions with no record at all. Where a number isn&#x27;t confirmed against rule text, the guide says so instead of estimating.</description>
  </item>
  <item>
    <title>OSS Protector — Flags AI-generated spam pull requests before they hit a maintainer&#x27;s review queue.</title>
    <link>https://stillworks.dev/products/p/oss-protector/</link>
    <guid>https://stillworks.dev/products/p/oss-protector/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>OSS Protector is a free GitHub App that scores incoming PRs on account age, diff signature and cross-repo overlap, then posts a public, auditable confidence breakdown for every flag. It&#x27;s community-run and free forever, with the detection weights open on GitHub for anyone to check. Coverage is still early: 348 flagged accounts tracked, live in 1 protected repo so far.</description>
  </item>
  <item>
    <title>VectorPop — Retraces PNG/JPEG logos into clean SVG locally on Windows and Linux, no uploads.</title>
    <link>https://stillworks.dev/products/p/vectorpop/</link>
    <guid>https://stillworks.dev/products/p/vectorpop/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>VectorPop turns a blurry raster logo into an editable SVG, vector PDF or high-res PNG entirely on your own machine — nothing is uploaded. It offers three presets (flat logo, detailed logo, black &amp; white line art) with a live preview before export, plus auto-tune and batch processing on the €39 one-time Pro tier. The free tier vectorizes unlimited images but caps you at 3 SVG exports a day.</description>
  </item>
  <item>
    <title>Spenco AI: Expense Tracker — Forecasts your month-end balance and EMI/SIP dates from forwarded bank alert emails.</title>
    <link>https://stillworks.dev/products/p/spenco-ai-expense-tracker/</link>
    <guid>https://stillworks.dev/products/p/spenco-ai-expense-tracker/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>Spenco parses the bank alert emails you forward from HDFC, ICICI, Axis, Kotak, Canara, PNB or Bank of Baroda into a live burn-rate forecast: days until your balance runs out, projected month-end spend, and category breakdowns. It&#x27;s read-only and forward-only — it never touches your inbox or bank login — and makes money from a ₹499/year subscription after a free first year, not from your data. Built specifically for Indian money constructs: EMIs, SIPs, UPI, salary cycles.</description>
  </item>
  <item>
    <title>LeadMatch — Scores each new HubSpot lead against reps&#x27; win history and picks who should take it.</title>
    <link>https://stillworks.dev/products/p/leadmatch/</link>
    <guid>https://stillworks.dev/products/p/leadmatch/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>LeadMatch reads up to 24 months of closed deals in HubSpot and ranks the top three reps for each new lead by expected value, posting the recommendation with its underlying win-rate math right in the HubSpot deal timeline. One click assigns the lead and its open tasks to the chosen rep, and a weekly load-balancing pass keeps top closers from getting everything. Setup is OAuth plus picking pipelines and attributes — about ten minutes, no code.</description>
  </item>
  <item>
    <title>tonhowtf/omniget — Downloads courses, videos and books from 1,800+ sites into one free desktop app.</title>
    <link>https://stillworks.dev/products/p/tonhowtf-omniget/</link>
    <guid>https://stillworks.dev/products/p/tonhowtf-omniget/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>OmniGet is a free, open-source (GPL-3.0) desktop app for Windows, macOS and Linux that downloads Udemy and Hotmart courses, YouTube/Instagram/X/Pinterest/TikTok video, and books, then plays, reads or transcribes them in the same window. It runs on yt-dlp and FFmpeg under the hood, installing and updating them automatically so there&#x27;s no terminal or flag cheat sheet to manage. No account, no ads, no telemetry on what you download.</description>
  </item>
  <item>
    <title>AlertiGate – Kubernetes alerts in, evidence-backed root cause out — Investigates Kubernetes alerts across metrics, logs and deploys, reports root cause to Slack.</title>
    <link>https://stillworks.dev/products/p/alertigate-kubernetes-alerts-in-evidence-backed-root-cause-o/</link>
    <guid>https://stillworks.dev/products/p/alertigate-kubernetes-alerts-in-evidence-backed-root-cause-o/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>AlertiGate is a self-hosted, read-only container that reacts to Alertmanager or Grafana alerts by having five agents read your Prometheus metrics, Loki logs, Kubernetes cluster, Argo CD deploys and your own runbooks, then posts an evidence-backed root-cause report to Slack before you open a terminal. It only ever runs GET/list calls — a test fails the build on anything else — and any fix it proposes is a diff you review, never one it pushes. Each investigation costs about 60k tokens under a ceiling you set.</description>
  </item>
  <item>
    <title>X-Radar — Alerts you within 2 minutes when a keyword matches on X, so you can reply first.</title>
    <link>https://stillworks.dev/products/p/x-radar/</link>
    <guid>https://stillworks.dev/products/p/x-radar/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>X-Radar scans X&#x27;s public search for your keywords every 90 seconds and pushes a Telegram alert with a median 46-second delay, aiming to get your reply in before X&#x27;s 30-60 minute ranking window closes. It supports X&#x27;s full search syntax, including min_faves: (not available on X&#x27;s own API), and shows exactly why a quiet period is quiet — quota, a broken scan, or Telegram blocked — instead of going silent. Pricing starts at $20/month for one keyword scanned every 90 seconds.</description>
  </item>
  <item>
    <title>VoxCut — Cuts silent pauses out of audio and video locally, no upload, one-time license.</title>
    <link>https://stillworks.dev/products/p/voxcut/</link>
    <guid>https://stillworks.dev/products/p/voxcut/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>VoxCut detects and removes silent gaps from podcasts, interviews and video locally, streaming the file through FFmpeg instead of loading it into RAM, so it handles files past 5GB — a 72-hour, 5GB session processed in under 6 minutes in their own test. It shows which segments it will cut before committing, so a quiet breath isn&#x27;t mistaken for dead air, and exports an XML/EDL timeline for Premiere or DaVinci on Pro. Free tier processes 10 minutes per file with a partial export, not a hard block.</description>
  </item>
  <item>
    <title>VectoFix — Repairs damaged zones of an already-vectorized image with a single brush stroke.</title>
    <link>https://stillworks.dev/products/p/vectofix/</link>
    <guid>https://stillworks.dev/products/p/vectofix/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>VectoFix re-rasterizes its own SVG trace and compares it pixel by pixel against your original, highlighting exactly where the vectorization drifted, then re-traces just that zone when you paint over it — cutting 60-80% of the local error in under a second. It&#x27;s explicitly not a general vectorizer: it&#x27;s built to repair an existing trace, yours or another tool&#x27;s, not to be a first-stop raster-to-vector converter. Runs 100% locally for €39 one-time, with a full trial that only locks export resolution.</description>
  </item>
  <item>
    <title>rabbitgui.com — GUI to debug, monitor and manage RabbitMQ with visual routing and drain-time forecasts.</title>
    <link>https://stillworks.dev/products/p/rabbitgui-com/</link>
    <guid>https://stillworks.dev/products/p/rabbitgui-com/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>RabbitGUI is a desktop client for RabbitMQ that lets you search and republish dead-lettered messages, spy on live queue traffic without consuming it, and forecast queue drain time with linear regression on historical backlog data. It handles multiple RabbitMQ environments — local, staging, production — from one interface with credentials stored locally, and includes a policy debugger that shows exactly which policies apply to a queue and why. A visual routing graph maps exchanges, queues and dead-letter paths instead of RabbitMQ&#x27;s plain management UI.</description>
  </item>
  <item>
    <title>TeCluster for Tesla — Turns your iPhone into a live Tesla instrument cluster over Bluetooth.</title>
    <link>https://stillworks.dev/products/p/tecluster-for-tesla/</link>
    <guid>https://stillworks.dev/products/p/tecluster-for-tesla/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>TeCluster mirrors your Tesla&#x27;s own speed, gear, battery and regen data onto your iPhone via a direct Bluetooth link, so the display keeps working in tunnels and parking garages where GPS/LTE-based apps drop out. It auto-starts navigation the moment you set a destination on the car&#x27;s own screen, ends guidance when you arrive, and runs a timed after-blow fan cycle to clear A/C odor when you park. Maps and routing come from Apple Maps, with HERE used only to reroute you if you go off-route.</description>
  </item>
  <item>
    <title>ExplainViz — Turns a written research description into an editable, journal-ready scientific figure.</title>
    <link>https://stillworks.dev/products/p/explainviz/</link>
    <guid>https://stillworks.dev/products/p/explainviz/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>ExplainViz drafts graphical abstracts, conceptual frameworks and scientific posters from a text description of your finding, then lets you edit labels, layout and colors before exporting as vector SVG sized to a specific journal — Elsevier, Wiley and Cell Press are named. It&#x27;s built for a narrow set of figure types researchers actually need, not general image generation, and exports carry no watermark.</description>
  </item>
  <item>
    <title>TickClip buyer-first shopping decisions platform — Tells you to buy, wait or skip an Amazon product after checking price and review history.</title>
    <link>https://stillworks.dev/products/p/tickclip-buyer-first-shopping-decisions-platform/</link>
    <guid>https://stillworks.dev/products/p/tickclip-buyer-first-shopping-decisions-platform/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>TickClip checks a pasted Amazon link against 90 days of price history, marketplace review signals and seller reputation, then returns one of three verdicts — Tick (buy), Clip (wait) or Skip (avoid) — with the evidence and a confidence level attached. It takes no affiliate commission from retailer links and says no seller can pay for a Tick, positioning itself as a buyer-side check rather than part of the sales funnel. It covers Amazon today; Walmart and Macy&#x27;s are shown as not yet supported.</description>
  </item>
  <item>
    <title>MRRDock — Adds up your MRR from Stripe, RevenueCat, Paddle and 4 more into one menu-bar number.</title>
    <link>https://stillworks.dev/products/p/mrrdock/</link>
    <guid>https://stillworks.dev/products/p/mrrdock/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>MRRDock polls Stripe, RevenueCat, Paddle, Lemon Squeezy, Polar, Dodo Payments, Gumroad and any custom HTTPS endpoint every fifteen minutes, converts everything to your display currency at live rates, and shows the total in your Mac menu bar. API keys sit in the macOS Keychain, never a config file, and the app talks to payment platforms directly — there&#x27;s no MRRDock server in between. It&#x27;s free and MIT-licensed with 37 tests, and deliberately stops at &quot;what is my MRR right now&quot; — no cohorts, LTV or churn like Baremetrics/ChartMogul.</description>
  </item>
  <item>
    <title>JustVugg/colibri — Runs a 744B-parameter MoE model on a 25GB-RAM consumer machine, pure C, zero deps.</title>
    <link>https://stillworks.dev/products/p/justvugg-colibri/</link>
    <guid>https://stillworks.dev/products/p/justvugg-colibri/</guid>
    <pubDate>Mon, 14 Sep 2026 02:19:04 +0000</pubDate>
    <description>Colibrì is an open-source inference engine that treats VRAM, RAM and disk as one memory hierarchy, streaming model experts from disk so a 744B-parameter MoE model (GLM-5.2) can run resident in under 10GB RAM. Nine model families run today, each as a single C file sharing the same chat/serve/web front end. It&#x27;s a research platform, not a polished product: the README states experiments must earn their place through reproducible end-to-end measurements, with a hard guarantee that low memory never silently changes model precision.</description>
  </item>
  <item>
    <title>DemoKit — Records CLI demo videos and refuses to save the file if the feature didn&#x27;t work</title>
    <link>https://stillworks.dev/products/p/demokit/</link>
    <guid>https://stillworks.dev/products/p/demokit/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>DemoKit reads your routes, components and tests, plans a demo flow, then drives the real app in headless Chromium and films it in 4K. Three independent checks verify each step actually happened before it writes the file — no green checkmark for a feature that silently broke. The camera is Cap&#x27;s zoom-and-spring engine ported as-is, driven by the recorded clicks rather than a model guessing where to point.</description>
  </item>
  <item>
    <title>Viaduct — C4 architecture modelling tool that serves your diagrams to coding agents over MCP</title>
    <link>https://stillworks.dev/products/p/viaduct/</link>
    <guid>https://stillworks.dev/products/p/viaduct/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>Viaduct models a system across the four C4 levels — context, containers, components, code — and keeps docs, sequence diagrams and OpenAPI contracts attached directly to each element instead of in separate tools. A rename or new dependency shows up at every level automatically, and an MCP server lets Cursor or Claude Code query the live model instead of a screenshot pasted into chat. It runs free in the browser with no account, and exports at any time.</description>
  </item>
  <item>
    <title>Reply Argus — Drafts and auto-publishes App Store and Google Play review replies in 100+ languages</title>
    <link>https://stillworks.dev/products/p/reply-argus/</link>
    <guid>https://stillworks.dev/products/p/reply-argus/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>Reply Argus watches every App Store and Google Play review in one inbox and drafts an on-brand reply before you wake up, grounded in your past approved answers and your store listing and marketing page. You approve or let trusted rules auto-publish; it replies in whatever language the review was written in, translated both ways without a separate tool. It also groups similar complaints into a ranked, evidence-backed list of what to fix next, and alerts you in Slack, Telegram, Discord or a webhook before a bad review spreads.</description>
  </item>
  <item>
    <title>PlayCISO — Simulates live security breaches and board/interview scenarios to train for the CISO role</title>
    <link>https://stillworks.dev/products/p/playciso/</link>
    <guid>https://stillworks.dev/products/p/playciso/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>PlayCISO runs you through incident simulations modelled on 17 years of real breaches across 3 fictional companies, then makes you defend your calls to a board and sit a mock interview with country-specific regulatory questions (PCI, HIPAA, MAS, DPDPA, NCA). It also includes three AI-copilot canvases for threat modelling (STRIDE), architecture design (NIST/CIS/ISO, with Terraform export) and RFC design review (OWASP ASVS). There&#x27;s a free 2-minute scorecard, but the full product runs $9/week, $29/month or $99/year with no free tier.</description>
  </item>
  <item>
    <title>Itemlist - Inventory Tracker — Tracks household or business inventory by home, room and container on iPhone/iPad</title>
    <link>https://stillworks.dev/products/p/itemlist-inventory-tracker/</link>
    <guid>https://stillworks.dev/products/p/itemlist-inventory-tracker/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>Itemlist keeps a searchable record of everything you own, organized by home, room and container like boxes or shelves, and finds any item by typing its name. It works offline, syncs across devices via cloud storage on EU servers, and exports the full inventory to CSV for insurance claims. The free plan caps at 100 items; a full trial unlocks unlimited items and the barcode scanner for small-business stock tracking.</description>
  </item>
  <item>
    <title>Monitrova — Monitors uptime, SSL, DNS and content changes and only alerts on confirmed issues</title>
    <link>https://stillworks.dev/products/p/monitrova/</link>
    <guid>https://stillworks.dev/products/p/monitrova/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>Monitrova checks uptime, SSL expiry, domain expiry, DNS changes, content changes and heartbeat pings for cron jobs, all on one timeline per site so you see the cause of an incident, not just the symptom. Alerts pass through a decision pipeline that defers non-critical issues for a second scan, bundles flapping issues into a daily digest, and logs every decision (sent, suppressed, deferred) with a reason — so a 3am false alarm doesn&#x27;t fire. Built by one person who runs it on his own 16 sites, plans start free for one site and run €5.99 to €49.99/month for 3 to 200 sites.</description>
  </item>
  <item>
    <title>NearHalo — Shares wellness, location and fall alerts between care recipients and family caregivers</title>
    <link>https://stillworks.dev/products/p/nearhalo/</link>
    <guid>https://stillworks.dev/products/p/nearhalo/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>NearHalo shares heart rate, blood oxygen, activity and location between a care recipient and the family or caregivers they choose, with separate warning and red-alert thresholds and a shared history of past alerts and outcomes. A companion app lets the care recipient control what&#x27;s shared and pause sharing at any time from phone or Apple Watch. It states plainly that fall detection requires supported devices, won&#x27;t catch every fall, and that NearHalo is not a medical device or emergency service.</description>
  </item>
  <item>
    <title>AgentMail — Gives AI agents real email inboxes via a REST API, built for autonomous workflows</title>
    <link>https://stillworks.dev/products/p/agentmail/</link>
    <guid>https://stillworks.dev/products/p/agentmail/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>AgentMail lets you create, send, receive and search email inboxes for AI agents through an API, with threads, attachments, realtime events, custom domains and semantic search rather than adapting a human inbox. It&#x27;s positioned specifically for agent use cases: extracting 2FA codes for browser agents, executive-assistant scheduling, parsing attachments, and routing support email. It has delivered 100M+ emails and offers instant inbox creation with no domain verification wait, and a free tier with no credit card required.</description>
  </item>
  <item>
    <title>TeaTonight: Dinner Planner — Weekly family dinner planner that sends the shopping list to Reminders, no tracking</title>
    <link>https://stillworks.dev/products/p/teatonight-dinner-planner/</link>
    <guid>https://stillworks.dev/products/p/teatonight-dinner-planner/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>TeaTonight plans seven days of dinners from the meals your family already cooks, with quick slots for dine-out, takeaway or leftovers, then pushes the full week&#x27;s ingredients to Apple Reminders as one consolidated shopping list. You can ask Siri what&#x27;s for dinner, and it sends exactly one daily reminder — no feed, no calorie counting, no ads, no account to create. The core app is free forever; Plus adds smart recipe import and multi-week planning for £19.99/year after a 7-day trial.</description>
  </item>
  <item>
    <title>Akta — Pay-as-you-go API for company data and news signals, built for AI agents</title>
    <link>https://stillworks.dev/products/p/akta/</link>
    <guid>https://stillworks.dev/products/p/akta/</guid>
    <pubDate>Sun, 13 Sep 2026 02:56:40 +0000</pubDate>
    <description>Akta gives structured data on 20M+ companies (70+ data points each) plus real-time and historical news signals, filtered to remove about 80% of noise, accessible via API, MCP, CLI or bulk data. Its own benchmark against six providers scores it highest on accuracy (93%) and F1 among News APIs, LLMs and scraper alternatives at $0.50 per 1,000 accurate articles, versus $0.67–$2.85 for competitors. Pricing is pay-as-you-go instead of the flat enterprise licenses legacy data vendors charge, and it states plainly it does not use customer data for AI model training.</description>
  </item>
</channel>
</rss>
