KB Report: Topics of the Last Month

Query: What have been my topics in the last month?
Date: 2026-04-24
Sources used: wiki/index.md + 18 wiki pages
Period covered: March–April 2026

Overview

Your clipping and research activity over the last month clusters into five clear theme areas, with one dominant thread running through almost everything: building AI-powered workflows and tools, particularly around Claude Code and Obsidian. Alongside that, you've been actively thinking about your own product and career direction, keeping tabs on Swift development, following a handful of stocks and entertainment, and reading one classic novel.

🧠 Theme 1: Claude Code + AI Workflows (Dominant — 10 sources)

This is by far your most active area. You've been deep in the emerging ecosystem around using Claude Code as an agent that reads and writes your filesystem — not just as a chatbot.

What you captured:

The full LLM Wiki pattern (Karpathy's idea): raw/ → wiki/ → reports/, where the LLM compiles and maintains a persistent knowledge base instead of re-deriving answers from scratch each query
The JARVIS setup: a full vault architecture with capture types (observations, reactions, patterns, questions, numbers), connection-finding skills, brief generation, and a voice-matched writing skill
Claude-Mem plugin: persistent cross-session memory via SQLite + background worker — solves the stateless session problem at the tool level
kepano's obsidian-skills: official SKILL.md files for Obsidian-native file formats (wikilinks, bases, canvas)
Karpathy's four coding guidelines: Think Before Coding, Simplicity First, Surgical Changes, Goal-Driven Execution — derived from his observations on LLM coding pitfalls
Superpowers plugin: 5-phase structured workflow (clarify→design→plan→code→verify), 9% cost / 14% token savings on complex tasks
Addy Osmani's agent-skills: production-grade SKILL.md library covering spec, TDD, code review, security, shipping

The thread connecting them: You're not just consuming content about AI tools — you're actively building a system using those tools to organise your own knowledge. The vault you're using right now (raw/ → wiki/ → reports/) is a direct instantiation of what you've been reading about.

📱 Theme 2: iOS App Idea + Career Direction (3 sources)

Running in parallel with the AI workflow research, you've been thinking seriously about a product and a career pivot.

The app idea: A native iOS second-brain app — SwiftUI frontend, Hummingbird backend — that brings the Obsidian + Claude workflow to iPhone. Core features: Markdown vault (local, iCloud-synced), screenshot → OCR → structured MD note, AI diary/memories via user-provided API keys, local vector store for RAG. Business model: 14-day trial, $5–10/month Pro.

The career angle: An ex-Meta engineer's framework for 2026 — use AI tools to build and ship polished apps, document publicly, land automation contracts. The key insight: engineering experience is still valuable, not for writing code, but for knowing what good output looks like and guiding the AI.

The connection: These two sources are pointing in the same direction — you're evaluating building something real with AI, not just using AI to read faster.

🦅 Theme 3: Swift Development (5 sources)

A steady background interest, less intense than the AI workflow theme but consistent:

Cross-compilation to Linux: build Swift on Mac with Static Linux SDK, deploy tiny Docker-free static binaries via Apple's Container CLI (new in 2025)
SwiftLint: two sources — the tool itself and its default rule configuration
Swift WASM: a thread on WebAssembly support (stub, not yet expanded)
iOS coding interviews: LeetCode-lite patterns — data stream aggregation, hashmaps, time/space complexity discussion

Observation: The cross-compilation source connects directly to the iOS app idea (Hummingbird backend would benefit from this deployment pattern). These aren't isolated interests.

📈 Theme 4: Stocks (1 source)

Aduro Clean Technologies (ADUR): Hydro Chemolytic Technology — a platform that breaks carbon-carbon bonds. Key insight: paraffinic (waxy) crude oil and polyethylene plastic are chemically near-identical (both are hydrocarbon chains, differing only in length). Aduro filed a patent for upgrading Utah Uinta Basin waxy crude into pipeline-flowable liquid without losing refining quality. The technology is the same as their plastic recycling work, just applied at a shallower "cut depth."

🎬 Theme 5: Entertainment (2 sources)

Light passive consumption, not a deep research area:

Project Hail Mary (film, 2026): Ryan Gosling sci-fi hit directed by Phil Lord & Christopher Miller. $420M worldwide in 2 weeks, 95% on RT. Surpassed Interstellar's domestic gross.
Beef Season 2 (Netflix, April 16, 2026): Anthology format, new cast — Oscar Isaac, Carey Mulligan, Charles Melton, Cailee Spaeny. Four-way blackmail battle at a country club.

📚 Theme 6: Reading (1 source)

Moby-Dick: Currently reading. Your note: slow in the middle (ship life and whaling descriptions), but enjoying the characters and Ishmael's perspective. Densely written.

📊 Activity Map

| Theme | Sources | Depth | Status | |---|---|---|---| | Claude Code + AI Workflows | 10 | Deep research | Active system being built | | iOS App + Career | 3 | Strategic thinking | Idea stage | | Swift Development | 5 | Moderate | Ongoing background interest | | Stocks | 1 | Moderate | Single company tracked | | Entertainment | 2 | Light | Passive | | Books | 1 | Light | In progress |

💡 Connections Worth Noting

Swift + iOS App + AI Workflows all converge: you're thinking about building an iOS app in Swift, using AI workflows, with a Hummingbird backend deployable via cross-compilation. These three themes are not separate — they're one potential project.
The meta-loop: You've been reading about building AI second brains while building one. The wiki you're reading this report from is proof the system is already working.
Career signal: The combination of AI workflow mastery + iOS/Swift skills + the "build apps with AI to land contracts" framework suggests you're positioning toward independent product work or AI consulting — not a traditional engineering job search.

Sources: wiki/claude-code-obsidian-second-brain.md, wiki/jarvis-obsidian-claude-code-setup.md, wiki/claude-mem-plugin.md, wiki/obsidian-skills-for-claude.md, wiki/karpathy-claude-code-guidelines.md, wiki/superpowers-plugin-claude-code.md, wiki/agent-skills-addyosmani.md, wiki/llm-wiki.md, wiki/ios-app-ai-second-brain.md, wiki/software-engineering-career-2026.md, wiki/business-model.md, wiki/swift-cross-compilation-linux.md, wiki/swift-exercise.md, wiki/swift-lint.md, wiki/swift-wasm.md, wiki/aduro-hydro-chemolytic-technology.md, wiki/project-hail-mary-film.md, wiki/beef-season-2.md, wiki/moby-dick-or-the-whale.md

kb-report-20260424-topics-last-month