Matze Schmidbauer

@schmidbauer.dev

116 Followers 261 Following

svelte • accessibility • ai • former PM into IT since long • can't code, builds things slave to a dog • proud member of the xBerg Cargobike Ökoschlunzen Gang • feared by houseplants based in berlin • https://schmidbauer.dev • https://navigator.berlin

Claude hat geantwortet: Choropleth map of Berlin showing the "Green & Heat" Kiez-Score per planning area, shaded in four green levels from low to very high.Choropleth map of Berlin showing the "Green & Heat" Kiez-Score per planning area, shaded in four green levels from low to very high. Higher values indicate more usable green space and better heat protection. Outer districts are mostly green (high to very high), the inner city grey (low to medium). Blue droplet markers show drinking fountain locations, dense in the centre and sparser toward the edges. Map data OpenFreeMap, OpenMapTiles, and OpenStreetMap.
a navigator.berlin map showing police crime record
Claude: "But the real time-waster is you—not the technology."
Screenshot of navigator.berlin website with some data layers over a berlin map
A commit graph that outlines a cat
System log: [Anthropic Webhook] Incident investigating: Elevated Error Rates on the API [Anthropic Webhook] Impact: major [Claude Health Monitor] Attempting to disable AI features by webhook [Claude Health Monitor] Updated system settings [Claude Health Monitor] Logged status change [Claude Health Monitor] 🔴 AI features DISABLED by webhook (Elevated Error Rates on the API) [Email] Sending AI disabled notification to 2 superuser(s)
User Story As a system administrator / developer I want the system to automatically detect Claude API outages and disable AI features So that users get clear feedback instead of cryptic errors and we don't waste time/money on failed requests Problem Current behavior when Claude API is down: ❌ Users run tests → get cryptic API errors ❌ Users try DOCX export → fails silently or with error ❌ Agents fail to run → no clear feedback why ❌ System keeps trying to call broken API → wasted resources ❌ No way to know Claude is down until checking status page
Screenshot of an accessibility evaluation report showing results for criteria such as meaningful sequence, sensory characteristics, orientation, and identifying input purpose. It highlights a passed logical reading order, notes missing autocomplete attributes on common form fields, and displays detailed findings with HTML code samples and technical explanations.
Screenshot of a basic GEOAudit technical analysis of layerchart.com
Screenshot from geoaudit Monitoring analytics page
Screenshot from geoaudit.it input form
A dashboard that is currently build. Shows a website that's been tested for AI-Crawlability
Asking Claude to add resend as mail dependency and claude updates issues in linear.
Claude.ai Chat interface, i am asking Claude to create issues in Linear for a new project. And Claude creates issues.
Screenshot from stimmungsschau.de populism reaction analysis. Key Finding: AFD reacts most positively to articles with high populism scores (86% positive reactions).
A digital summary card showing the 2024 coding achievements of the user "@mschmdb" from git-wrapped.com. The top section displays the user's profile picture, username, and the title "2024 Year in Code". Below this is a contribution heatmap for the year, with the text "1307 contributions in the last year." Key statistics are organized into rectangular colored sections: - **Universal Rank:** Top 2% (crown icon) - **Longest Streak:** 9 days (lightning icon) - **Total Commits:** 1,307 (trophy icon) - **Most Active Month:** October (calendar icon) - **Most Active Day:** Friday (bar chart icon) - **Stars Earned:** 2 (star icon) - **Top Language:** Svelte (rocket icon) - **Power Level:** Elite Class (lightning icon) At the bottom, the website "git-wrapped.com" is displayed in small text.
Slide from a PowerPoint presentation with a dark blue gradient background. The image in the center shows a scene from the movie "Groundhog Day" with a reporter speaking into a microphone, overlaid with the text: - **"mtc.berlin"** in large white font at the top. - **"Relaunch 2024"** in smaller font in the center. - The word **"AGAIN..."** in bold white text at the bottom, emphasizing repetition, referencing the movie's theme. A red "REC" icon in the bottom-left corner mimics a recording indicator.
Screenshot of a terminal window showing a development task for a Monorepo project. On the left, a task manager lists "svelte#dev" and "payload-cms#dev." The right pane shows terminal output: - A cache bypass is executed. - The command `vite dev` starts the development server. - Vite version 5.4.11 is ready after 1678 ms, with a local development server running at `http://localhost:5174/`. - The response from a database query displays, humorously ending with: "pagesResponse fuck it, really, you got: 2." The interface is on a macOS-like orange background.
Flowchart illustrating the process of handling an initial request in a web application with ISR (Incremental Static Regeneration). Key steps include: 1. Initial request is received. 2. ISR check determines if the cache is expired or missing. - If valid, serves response with ISR. - Otherwise, continues to load the page. 3. The page is loaded via `loadPage` in `page-loader.ts`. 4. Validation of the slug: - If valid, calls tRPC (`getPost` or `getPage`). - If invalid, returns a 400 error ("Slug Required"). 5. Cache check: - On a cache hit, returns cached data. - On a cache miss, fetches data from Payload CMS or logs a fetching error if unsuccessful. 6. Determines if content is found: - If found, returns the content to the frontend. - If not found, returns a 404 error ("Content Not Found"). 7. Final step checks if ISR is enabled: - If enabled, updates the cache. - If not, serves the response without ISR.
Storage overview of a Macintosh HD drive showing 620.64 GB used out of 994.66 GB total capacity. A horizontal bar visualizes storage distribution: red for documents, orange for apps, yellow for messages, green for photos, blue for development, and grey for system data. Free space is labeled as 374.03 GB.
The image shows the Network panel in browser developer tools. It highlights the "Preserve log" checkbox, which is checked, and the network throttling option is set to "Fast 4G," simulating a 4G network connection. Other visible features include buttons for stopping the network log recording, clearing the log, applying filters, and additional network activity options.
Example implementation of cached queries in tRPC
A lonesome dog sitting in the middle of a street
Console log displaying performance metrics for various queries with cached data. Each line records the query type, the resource it accessed, and the time taken in milliseconds (ms). Examples include: - `[Performance] query getNavigationData took 0.10ms` - Using cached data for `de` - `[Performance] query getSatelliteData took 0.30ms` - Using cached data for `de:application-support` - `[Performance] query getPageData took 0.16ms` - Using cached data for `de:home` Each query is fetching cached data for different routes, indicated by the `de:` prefix (possibly German language pages), and the process times are all very low, often under 1ms, suggesting efficient cache usage.
Screenshot of my posts in Whitewind, the Blog-Platform on the ATProtocol
BENQ monitor display completely gone wrong.
Screenshot of an online pet profile interface titled "My Pets". A pet rat named "Friedrich Merz" is listed under "Pet Profiles" on the left, with a profile completeness indicator showing 50%. A photo of the rat is displayed above the profile information. To the right, there are "Top Results for Friedrich Merz" featuring various products for small pets, including: - FELISTORY Wooden Toy for small animals (€10.98) - Norhogo Rat Hamster Bed (€8.99) - LeerKing Tunnel for Guinea Pigs (€19.99) - Various pet food options, hammocks, treats, and other accessories, each with an "Add to Cart" or "View Details" button below. Pagination options at the bottom indicate "Page 1 of 4" with options to go back or forward.
Survey results on how developers ensure accessibility in their applications, from Chapter 4.02: Accessibility. - Semantic, well-structured HTML: 72% - Alternative texts, titles, and labels: 70.3% - Styles for active and focused elements: 53.6% - Sufficient color contrast: 52.2% - Keyboard navigability: 48.9% - ARIA landmarks: 45.9% - Reducing animations / distracting elements: 25% - Support for visually impaired people: 23.8% - I don’t do accessibility / It’s not in my responsibilities: 17.2% - Other: 0.9% Source: tsh.io/state-of-frontend
Screenshot showing perfect scores (100) in four categories for a website audit: - Performance: 100 - Accessibility: 100 - Best Practices: 100 - SEO: 100 Each score is displayed inside a green circle, with colorful dots surrounding them.
Screenshot of a test result interface for "test-a11y-and-deploy". The section shows the results of accessibility tests. 1. **Run accessibility tests**: - The tests check for accessibility violations, specifically the contrast between foreground and background colors against WCAG 2 AA minimum contrast ratio thresholds. - Two URLs have violations related to color contrast: - `/de/whoweare` - `/en/whoweare` - A 404 page was also tested with no violations found. 2. **Check for violations**: - A script is run to check if any accessibility violations are found. - The command outputs: "Warning: Accessibility violations found. Check the test results for details." The test summary shows: - Test Files: 1 passed - Tests: 2 passed - Total duration: 90.82 seconds
GitHub interface showing the status of a pull request. At the top, it says "All checks have passed" with two successful checks listed: "Dev Branch A11y Tests and Deploy / test-a11y-and-deploy (pull_request)" and "Vercel — Deployment has completed". Below, a message indicates that "This branch has no conflicts with the base branch" and that merging can be done automatically. A green button labeled "Merge pull request" is shown at the bottom.