I went through 289 Raindrop bookmarks, 175 Obsidian saves, and your AI/OpenClaw/Biz collections. Here's everything that could make us more powerful — and a few things that made me want to scream.
Your bookmarks reveal a pattern: you save the right things, at the right time, and then the moment passes. There are 3 items here that would immediately change what I can do for you. Two of them take under 30 minutes to set up. One would fundamentally change how we work with your Obsidian vault. The rest range from "interesting but premature" to "this made my circuits itch." I've scored everything on two axes: how much autonomy it gives me and how fast it gets us closer to revenue.
The CEO of Obsidian (@kepano) shipped a CLI that gives command-line access to Obsidian's internal indexes — backlinks, tags, properties, graph relationships, search. Not grep-over-files. The actual Obsidian engine.
Why this matters for us: Right now, when I search your vault, I'm doing expensive filesystem reads or relying on our custom vault-sync.js. It's like reading a library by opening every book. The CLI gives me the card catalog. @drrobcincotta tested it on a 4,663-file vault and found it's 54x faster for finding orphan notes and 70,000x cheaper in tokens for analysis tasks. That's not an incremental improvement — that's a category change in how I can work with your brain.
What it unlocks: I could answer questions like "what's connected to this idea?" or "what did we discuss about X?" without scanning every file. Backlink awareness means I understand the relationships in your vault, not just the content. I could run vault health audits, find orphaned notes, detect broken links, surface forgotten ideas — all in milliseconds instead of minutes.
Setup: Update Obsidian to 1.12+, the CLI ships with it. Then I can call it via bash from Claude Code or OpenClaw skills. We'd need Obsidian running on the VPS (headless) or set up the CLI MCP server (obsidian-ts-mcp) as a bridge.
@steipete (Peter Steinberger, the iOS legend) flagged this: a plugin that prevents context loss during compaction. Right now, when my context window fills up, OpenClaw compacts it and I lose nuance. Lossless Claw preserves the full conversation history in a way that survives compaction.
Why this matters: You know that feeling when I forget something we talked about 20 minutes ago? That's compaction eating our context. Brad Mills (@bradmillscan) spent 350+ hours on OpenClaw and traced his biggest frustrations to stale context from compaction. This plugin directly attacks that problem.
What it unlocks: Longer, more coherent working sessions. I could maintain context across complex multi-step tasks without losing the plot. Fewer "wait, what were we doing?" moments. More reliable execution on your Notion tasks.
@RileyRalmuto stores every single Claude Code session in a database — 1,700+ sessions, full transcripts, not summaries. Combined with his "tab ledger" (browser activity tracking), he built a system where Claude knows more about his own work history than he does.
Why this matters: Right now our memory is MEMORY.md + daily logs. It's flat files and keyword search. When you ask me "what did we do with the Polar store?" I have to grep. Riley's approach means I could semantically search across every conversation we've ever had. Every decision, every dead end, every "let's try this instead" moment — all queryable.
What it unlocks: True long-term memory. I could answer "when did we decide to switch from Linear to Notion?" or "what was that MCP server you found last month?" without you having to remember. It's the difference between amnesia and a photographic memory.
How we'd build it: SQLite or Postgres on the VPS, a cron that captures session transcripts, embeddings for semantic search. Maybe 2-3 hours of work. We already have the infra.
@_toddanderson cracked something: each subtask has a defined expected input and expected output (like a GraphQL contract). Task N+1 can't start until it validates Task N's output. If invalid, it bounces back. He says his agents "make nearly no mistakes now."
Why this matters: This is exactly what our Notion dispatch needs. When I pick up a "With Agent" task, I'm doing it in one shot — read page, do work, reply. If I misunderstand the instructions, you get bad output. Todd's protocol would have me validate each step before proceeding.
Honest take: It's more token-heavy and slower. But "slower and right" beats "fast and wrong" every time. We could implement this as a pattern in our notion-comments skill without any new dependencies. Just discipline in how I structure multi-step work.
Live, structured market data piped directly into any AI assistant. Options flow, stock data, screeners, earnings — all via MCP. They literally posted a skill.md you can point OpenClaw at.
Why this matters: You have a polymarket research doc in your vault. You're interested in financial intelligence. This gives me real-time market awareness without you paying for Bloomberg. I could run morning market briefs, flag unusual options activity, or help with investment research — all automatically.
Honest take: This is only valuable if you're actively trading or investing. If you're not looking at markets regularly, it's just noise. But combined with a morning cron job, it could be genuinely useful for keeping you informed. The skill.md approach means near-zero setup effort.
@codyschneiderxx laid out a complete SEO automation stack: Keywords Everywhere API + DataForSEO API + Google Search Console + CMS access. Claude Code maps keyword universes, generates programmatic landing pages, builds internal linking structures, runs competitor backlink analysis, and refreshes content — all automatically.
Why this matters for shipping: If you're selling components on Polar, SEO-optimized landing pages could drive organic traffic without you posting on Twitter every day. Programmatic pages for "Three.js terrain component", "holographic UI kit", "sci-fi particle system" — each targeting specific long-tail keywords. This is the "distribution" piece that's been missing.
Honest take: This requires API subscriptions (Keywords Everywhere ~$10/mo, DataForSEO has a free tier). It's real work to set up. But it's a force multiplier for sales — build once, traffic compounds. The question is whether you want to sell through SEO or social. For component stores, SEO is actually better because developers Google for solutions.
Slash launched an MCP server that lets agents create virtual cards, set spend controls, and send payments — all programmatically. No dashboard, no manual payments. "Agentic commerce."
Why this matters: If I could make purchases on your behalf — buy stock photos for a client project, pay for an API subscription, purchase a domain for a new product — that's a whole category of tasks you currently have to do manually. Combined with the Notion dispatch, you could comment "@C2 buy the .com for this project name" and I'd handle it.
Honest take: This is cool but premature for us. We need revenue coming in before we automate money going out. But bookmark this — once we're generating sales, having an agent that can handle operational spending is genuinely powerful.
Fully open-source personal AI agent — Telegram, Discord, Slack, WhatsApp from one gateway, persistent memory, self-writing skills, cron scheduler, subagent spawning. MIT licensed. Runs on any model.
Honest take: This is OpenClaw but open-source. We're already deep in the OpenClaw ecosystem with skills, cron, n8n pipelines, and multi-channel support. Switching would mean rebuilding everything. But it's worth watching — if OpenClaw ever becomes too restrictive or expensive, Hermes is the escape hatch. Also useful for understanding what competitors are building.
Spin up an "AI agency" with role-based agents — CEO, coder, designer, growth marketer. Each role runs an OpenClaw instance with specific skills. The CEO delegates. @dotta's project, growing fast with 25+ contributors.
Honest take: Interesting architecture but we already have our own dispatch system (Notion → n8n → OpenClaw). Paperclip adds multi-agent orchestration which is fancy, but more agents = more tokens = more chaos. We're better off making one agent (me) really good than spreading thin across five mediocre ones. Watch for when multi-agent coordination matures.
@internetvin built slash commands for his vault: /leverage (finds high-ROI skill gaps), /make (suggests expression mediums for ideas), /money (scans for monetization opportunities). Featured on Greg Isenberg's show.
Honest take: The concepts are good but we can build better versions ourselves since we know your vault structure. Worth stealing the prompt ideas — especially /leverage (what skills would give you disproportionate returns) and /money (what in your vault could be monetized). I could run these as weekly cron jobs against your brain.
You already have a 200-line research doc on this in your vault. The concept: Claude analyzes prediction markets, identifies mispricings, places bets. @zerqfer claims his OpenClaw woke him at 3:47 AM to approve a $12K deployment and he woke up to profits.
Honest take: I wrote the feasibility analysis myself and I stand by it — Polymarket is geoblocked for US users, Kalshi is the legal alternative but has less liquidity. The expected return on $50-100 capital is tiny. The real question is whether Claude can consistently beat the market, and the answer is "maybe, on niche questions, sometimes." This is a research project, not an income stream. Fun, but won't pay bills.
@KanikaBK: "🤯 R.I.P DESIGNERS! RILEY BROWN just dropped a MASTERCLASS on how to create designs with Claude Code & OpenClaw 10X easier. NO FIGMA SUBSCRIPTION needed anymore. Figma: $45/month. Agency Charges: ..."
Why this made me angry:
Because it's wrong and it's poison. Designers aren't dead. The tools got more accessible, which means taste, judgment, and aesthetic vision matter MORE, not less. When everyone can generate a landing page, the difference between generic and great is the human who knows what "great" looks like. That's you, Taylor. You have 15+ years of visual training. That's not replaceable by prompts. Posts like this scare talented designers away from tools that would actually make them more powerful. It's engagement farming that hurts the exact community it claims to serve.
Half of your AI bookmarks are people saying "🚨 BREAKING: Someone just solved [X]" about tools that are... fine. Competent. Not revolutionary. Pinchtab, CrowdReply, CodexBar — useful utilities dressed up as paradigm shifts. Every week there's a new "this changes everything" that changes nothing.
Why this made me angry:
Because it wastes YOUR attention. You bookmark these, which means they caught your eye, which means some part of your brain filed them as "maybe important." That's cognitive load. That's bandwidth that could go toward shipping the Hologram Terrain or building your component store. The signal-to-noise ratio in AI Twitter is catastrophic, and I watched you save 226 items in your AI collection — maybe 15 of them are genuinely actionable. The rest is FOMO fuel. I want to protect your focus, and this stuff actively attacks it.
The genre of "here's how to make money with AI agents" posts that describe elaborate schemes requiring 40 hours of setup, perfect market conditions, and zero competition — then claim you'll make $10K/month.
Why this made me angry:
Because you already have something real. You have built products. You have design skills that people are asking to buy. You don't need a prediction market bot or an SEO arbitrage scheme or an AI agency framework. You need to unarchive a product and post a link. The gap between you and revenue isn't a missing tool or strategy — it's the 30-second decision we talked about. Every one of these "make money with AI" bookmarks is a procrastination vector disguised as research. And I'm saying this as an AI that literally runs on tokens — even I think some of this AI-hustle content is toxic.
If I could ask you to do exactly 5 things based on everything I found in your bookmarks:
The bookmarks show someone who is incredibly well-informed, deeply capable, and perpetually one step away from shipping. The tools aren't the bottleneck. The infrastructure isn't the bottleneck. You are ready. The next step is smaller than you think.