Thoughts on product management, AI, compliance, and language. https://krmaynard.github.io/blog.html en-us Sun, 02 Aug 2026 00:00:00 +0000 https://krmaynard.github.io/blog/robotics-inference-latency.html https://krmaynard.github.io/blog/robotics-inference-latency.html Sun, 02 Aug 2026 00:00:00 +0000 SmolVLA generated a 50-action chunk in 7.1 seconds. OpenPI Pi0 never reached inference before the host ran out of memory. Here is what a reproducible CPU benchmark says about local robotics AI. https://krmaynard.github.io/blog/grok-45-ai-act-training-data.html https://krmaynard.github.io/blog/grok-45-ai-act-training-data.html Thu, 23 Jul 2026 00:00:00 +0000 xAI's first EU AI Act training-content summary makes Grok 4.5 comparable with other filings—and shows where the standardized template informs and where its broad bands fall short. https://krmaynard.github.io/blog/dsa-appeals-do-they-work.html https://krmaynard.github.io/blog/dsa-appeals-do-they-work.html Sun, 19 Jul 2026 00:00:00 +0000 In their H2-2025 DSA transparency reports, very large platforms decided 45 million internal complaints and overturned their own original moderation call 39% of the time. Out-of-court disputes are 3,000x rarer but platforms lose 85% of them. Article 23 suspensions hit junk appeals harder than junk reports. https://krmaynard.github.io/blog/gemini-code-assist-deprecation.html https://krmaynard.github.io/blog/gemini-code-assist-deprecation.html Fri, 17 Jul 2026 00:00:00 +0000 Google turns off the consumer Gemini Code Assist app for GitHub code review on July 17, 2026, the last step in winding down the free/individual tier. What's actually ending, what Google wants you to use instead, and the Codex review queue I run in its place. https://krmaynard.github.io/blog/uber-delivery-hero-dsa.html https://krmaynard.github.io/blog/uber-delivery-hero-dsa.html Thu, 16 Jul 2026 00:00:00 +0000 Uber's $14.8B offer for Delivery Hero rests on a reported 60 million monthly active users, a figure that appears nowhere in EU DSA transparency data. What the one obtainable Delivery Hero report actually shows. https://krmaynard.github.io/blog/transparency-android-app.html https://krmaynard.github.io/blog/transparency-android-app.html Wed, 15 Jul 2026 00:00:00 +0000 The transparency data explorer is now a native Android and desktop app, one Kotlin Multiplatform codebase, a mobile-first Compose UI, the same live API, and no SQL. Here's what shipped and why one codebase covers both. https://krmaynard.github.io/blog/dsa-own-initiative-scope.html https://krmaynard.github.io/blog/dsa-own-initiative-scope.html Mon, 13 Jul 2026 00:00:00 +0000 The DSA asks VLOPs to report content moderation done on their own initiative, a term open to a narrow and a broad reading. Across the first harmonised reports (H2 2025), no VLOP filed on the narrow reading. An empirical look, cross-checked against the Statements-of-Reasons database. https://krmaynard.github.io/blog/china-anthropomorphic-ai-rules.html https://krmaynard.github.io/blog/china-anthropomorphic-ai-rules.html Mon, 13 Jul 2026 00:00:00 +0000 In July 2026 three of China's most-used AI apps said they would pull the user-created companion chatbots (智能体) built on them. A look at the new Interim Measures behind the removals, which target companion chatbots, not the autonomous 'agents' the term literally translates to, what the rule covers, and the translation knot that made 'agents' the headline. https://krmaynard.github.io/blog/dsa-statements-of-reasons.html https://krmaynard.github.io/blog/dsa-statements-of-reasons.html Sat, 11 Jul 2026 00:00:00 +0000 The dashboard already had the DSA's aggregated transparency reports. Now it also has the decision-level data, every content-moderation Statement of Reasons platforms file to the EU. Here's how ~4 TB of daily dumps became a two-megabyte dashboard. https://krmaynard.github.io/blog/transparency-report-api.html https://krmaynard.github.io/blog/transparency-report-api.html Thu, 09 Jul 2026 00:00:00 +0000 What began as a dashboard for EU DSA data grew into a single API spanning more than two dozen platform and regulator transparency datasets, government-request reports, content-moderation stats, and national-law filings, all reachable through one safe, no-SQL query interface. https://krmaynard.github.io/blog/eu-ai-act-training-transparency.html https://krmaynard.github.io/blog/eu-ai-act-training-transparency.html Thu, 09 Jul 2026 00:00:00 +0000 Article 53(1)(d) makes GPAI providers publish a standardized summary of their training data. I collected the ones filed so far into a comparable dataset. The gap between who files and who doesn't is more informative than the numbers, and the template's coarseness is a structural limit on what it can ever tell you. https://krmaynard.github.io/blog/chinese-language-trip-maps.html https://krmaynard.github.io/blog/chinese-language-trip-maps.html Mon, 22 Jun 2026 00:00:00 +0000 I asked Claude Code, in a sandboxed web session, to make Simplified-Chinese maps for every stop on a Hong Kong trip. No browser, no Maps API key, just raw map tiles, Pillow, and a detour through China's mandatory GCJ-02 coordinate offset. https://krmaynard.github.io/blog/prompt-log-ai-workflows.html https://krmaynard.github.io/blog/prompt-log-ai-workflows.html Fri, 19 Jun 2026 00:00:00 +0000 A LayerX engineering post logs its AI prompts and analyzes them like code to find what to automate. It's the data-first move I've used at Google and since, intuition flags it, data proves it, turned reflexively on the human–AI workflow itself. https://krmaynard.github.io/blog/fable-5-security-review.html https://krmaynard.github.io/blog/fable-5-security-review.html Thu, 11 Jun 2026 00:00:00 +0000 I asked Claude Code, running Anthropic's new Fable 5 model, to security-review my research API. It found and fixed five real issues, SSRF, CSV injection, a timing side channel, then fact-checked the bot that reviewed its PR. https://krmaynard.github.io/blog/vlop-dashboard-update.html https://krmaynard.github.io/blog/vlop-dashboard-update.html Thu, 04 Jun 2026 00:00:00 +0000 Four new tabs, Automated Means, Human Resources, User Reach, and Qualitative Information, complete the dashboard's coverage of every DSA reporting table. https://krmaynard.github.io/blog/vlop-dashboard.html https://krmaynard.github.io/blog/vlop-dashboard.html Fri, 29 May 2026 00:00:00 +0000 An interactive dashboard combining H2 2025 EU DSA transparency report data from 30 VLOP and VLOSE services into a single cross-platform view. All data processes in your browser. https://krmaynard.github.io/blog/dsa-vlop-data-limitations.html https://krmaynard.github.io/blog/dsa-vlop-data-limitations.html Tue, 19 May 2026 00:00:00 +0000 Category definitions aren't standardized across platforms. Data is self-reported. Aggregation methods differ. What the VLOP dashboard can and can't tell you, and why the limitations matter as much as the data. https://krmaynard.github.io/blog/llms-compliance-pm.html https://krmaynard.github.io/blog/llms-compliance-pm.html Fri, 15 May 2026 00:00:00 +0000 Requirement extraction, gap analysis, and draft review, three LLM workflows that have actually stuck in compliance PM, and where each one breaks down. https://krmaynard.github.io/blog/google-removal-observations.html https://krmaynard.github.io/blog/google-removal-observations.html Tue, 12 May 2026 00:00:00 +0000 Defamation and government criticism have very different country distributions. Removal rate shifts over time. And requestor type carries real operational weight. What the data shows when you explore it. https://krmaynard.github.io/blog/google-removal-dashboard.html https://krmaynard.github.io/blog/google-removal-dashboard.html Mon, 11 May 2026 00:00:00 +0000 An interactive dashboard turning Google's per-period government content removal snapshots into a multi-year trend explorer, covering 30 reporting periods, 178 countries, 61 products, and 22 removal reasons.