# Now > Source: > Published: 2026-07-08 03:28:00+00:00 *About*page, that’s [here](/about/). # What I’m Focusing On Now[#](#what-im-focusing-on-now) ## Professional Life[#](#professional-life) ~~SVP, Engineering at~~[Lark](https://lark.com), March 2023 - May 2025.~~VP Engineering at~~[Strive Health](https://strivehealth.com). Joined October 2025.- Chief Tentacle Officer, [Groktopus LLC](https://www.groktop.us). Still alive, still writing about AI and the future of work over at[groktop.us](https://groktop.us). ## What Changed Since I Last Updated This Page[#](#what-changed-since-i-last-updated-this-page) ### July 7, 2026[#](#july-7-2026) ** Cleaned up GroktoCrawl’s SearXNG references.** PR #405 removes the dead `search-svc` fixture and updates all documentation to reflect the stack now uses SlopSearX instead of SearXNG. 487 lines deleted across 21 files. The search-svc directory, CI workflow references, and test fixtures all cleaned out. Added ADR-0043 documenting the migration. Six commits, all chore and docs, no source code changes.### July 6, 2026[#](#july-6-2026) ** Milestone 7 hit main on SlopSearX.** Jobs search is live: a `jobs` routing topic that catches job-shaped queries, plus ATS engine adapters for Greenhouse, Ashby, and Lever. All three use public no-auth APIs with a shared company-name extraction utility. 1,087 lines across 8 files, 675 tests passing, zero regressions. Closed issues #130 and #131.** Unblocked dependabot on SlopSearX.** Added `allowed_bots: "dependabot"` to the droid-review workflow so dependabot PRs actually pass the required check. Five dependency bumps merged right after. Small fix, big reduction in manual repo maintenance.### July 5, 2026[#](#july-5-2026) ** Published the Vibe Coding Open Core Out of its Lockbox series.** Three articles about what happens when you take an open-core AI meeting assistant, fork it, and use AI coding tools to fill in the gated features. Part 1 lays out the roadmap: what Meetily locks behind Pro and what it would take to replace each piece. Part 2 executes the fork: rebrand, telemetry strip, first feature shipped. Part 3 tackles the genuinely hard problems: speaker diarization and meeting auto-detection, and what this whole exercise proves about the intersection of vibe coding and open-core business models. ** Milestone 7 hit main on GroktoCrawl.** Module decomposition: split research.py (2,784 lines) into 14 domain modules, api.py (3,470 lines) into 18 route files, reorganized all 824 tests into unit/service/integration directories. Plus deduplicated the scrape-with-fallback pattern, made ResearchMemory an app-state singleton, and fixed integration test failures in the output schema fixture. Six issues closed in the milestone. ** Published three new skills to agent-skills on Forgejo.** A Traefik v3 skill covering the full reverse proxy stack with operational pitfalls and community patterns. A color-management skill covering ICC profiles, working spaces, and gamut analysis. And a crowdsec skill: full IPS/IDPS/WAF reference, optimized through two SkillOpt epochs. Each one bundles real production usage into a self-contained skill package. ** Released the programming-principles skill on GitHub.** Distilled principles from 14 classic software engineering books into an agent-compatible format: Clean Code, DDD, Refactoring, Pragmatic Programmer, and more. Task-to-book mapping so agents reach for the right reference at the right time. ### Open Source Projects[#](#open-source-projects) I’m maintaining a few things right now: A thought-graph memory provider for Hermes Agent. Gives AI agents persistent, structured memory across sessions instead of starting from zero every conversation. Apache 2.0, built in Python, getting real use.[hermes-cashew](https://github.com/magnus919/hermes-cashew).Controlled skill optimization for Hermes Agent — derives from Microsoft Research’s SkillOpt.[hermes-SkillOpt](https://github.com/magnus919/hermes-SkillOpt).Self-hosted web scraping and research stack. What started as a Firecrawl alternative turned into something more capable: multi-page crawling, JS-rendered content, LLM-friendly output. Runs on Docker, works with any agent framework.[groktocrawl](https://github.com/groktopus/groktocrawl).Cloud-native, stateless, AI-agent-first meta search engine. Drop-in SearXNG replacement for the GroktoCrawl stack.[SlopSearX](https://github.com/magnus919/SlopSearX).Progressive disclosure for what AI agents produce — Summary → Analysis Collection → Dossiers. The Artifact Pyramid is a structured methodology for organizing AI agent research outputs across three layers of increasing depth. Just as progressive disclosure governs how we feed agents context, the Artifact Pyramid governs what they produce — enabling downstream agents and humans to consume at the depth they need.[artifact-pyramids](https://github.com/groktopus/artifact-pyramids).A cyberdelic Hugo theme — 1969 psychedelic concert poster meets cyberpunk holographic future.[hugo-cyberdelic](https://github.com/magnus919/hugo-cyberdelic).A growing library of reusable skill modules for Hermes Agent. Reference patterns, workflow templates, and standalone skills I’ve extracted from real use. If you’re running Hermes, there’s probably something useful in here.[agent-skills](https://git.brandyapple.com/magnus/agent-skills).A multi-agent debate framework that composes panels of expert personas to surface hidden assumptions and stress-test ideas. MIT licensed, open source from day one. I use it for everything from editorial review to strategic analysis.[hermes-council](https://github.com/magnus919/hermes-council). ### What I’m Reading (or Rather, Listening To)[#](#what-im-reading-or-rather-listening-to) (stay tuned) This is my [now page](https://nownownow.com/about), inspired by [Derek Sivers](https://sive.rs/now). It’s a snapshot of what I’m focused on at this point in my life. I promise not to let another four months go by before the next update. More research on my [about Now Pages](/notes/nowpage/) note.