Mario Miletta

@wcools The only thing I'd like is to be able to use my Claude sub in Thymer without having to use the MCP

executable documents? run note.md

@wcools Looks great

bullish on the PM role quietly becoming the most important role in tech again when anyone can build, the person who decides WHAT to build becomes the bottleneck

@AnthropicAI should build the next Notion/Obsidian @trq212 @bcherny

@lucaronin Looks really nice 👏

Official release date for Tolaria is set for next Wed! Writing this in public to stay accountable! Now I am writing the getting started and wow getting proud of this 🥲

@petergyang Start a remote session via web app

I'm enjoying using the new Claude Code app (RIP terminal). How can I work on the same session across the Claude mobile app too? /remote-control doesn't seem to work and tbh I shouldn't have to type this to work across desktop + mobile anyway.

You can use AI to cut costs. You can also use AI to raise your ambition. Too many people focus on the first. More people should be thinking about the second.

The people building elaborate second brain setups with 400K words of wiki are optimizing for knowledge. I'm optimizing for shipping. Obsidian is the brain. Claude Code is the hands. Keep the context tight, and the output compounds.

My workflow: rough spec in Obsidian, Claude Code reads context and builds. I review, iterate, push. The gap between "I know what to build" and "it's live" went from days to hours. Not because the AI is smart. Because it has the right context to be useful.

The trap everyone falls into: dumping more notes thinking more context = better output. It doesn't. I learned this the hard way. What works: a curated CLAUDE.md file that tells the agent what matters, where to look, and what to ignore. Think onboarding doc, not brain dump.

I've been shipping product with this stack for months. Obsidian isn't my second brain. It's my product OS: specs, decision logs, user research, architecture docs. All markdown, all linked. Claude Code reads it and has the full context of what I'm building and why.

Karpathy uses Obsidian + Claude Code to compile knowledge bases. Fraser Cottrell turned it into an "AI employee." Both are missing the point. The real value isn't remembering. It's building faster.

You have no experience. You’ve never started a company. You’ve never had a full time job. Nike is going to kill you. You’re a kid. You don’t have technical skills. You shouldn’t build hardware. Apple is going to kill you. You can’t build hardware. You can’t measure heart rate non-invasively. Athletes don’t care about recovery. Under Armour is going to kill you. It won’t be accurate. You don’t listen. You’re an ineffective leader. You can’t recruit great talent. You’re going to have to pay every athlete. You can’t measure sleep non-invasively. It’s too expensive to research. Athletes are a small market. The product costs too much to make. The product costs too much to sell. Your valuation is too high. Consumers aren’t going to want it. Hardware is too hard. You should measure steps. Fitbit is going to kill you. You can’t build a marketing engine. You can’t raise enough money. You need a real CEO. Google is going to kill you. You can’t be a subscription. You can’t build a brand. You can’t do consumer in Boston. Your valuation is too high. You shouldn’t make accessories. You shouldn’t make apparel. Lululemon is going to kill you. You can’t predict Covid. Stay in your niche. You are going to run out of money. You can’t build a health platform. Amazon is going to kill you. You can’t measure blood pressure. You can’t get medical approvals. The market is too small. You don’t understand AI. The market is too competitive. It won’t work internationally. The supply chain is too complicated. You can’t build an AI. You can’t raise enough money. It’s too competitive. Healthcare isn’t going to want it. … Just keep going ✌️

@PawelHuryn So, Claude is better, but codex is cheap

"Usage of Codex is growing very fast." The chart has no Y-axis scale. OSWorld (computer use): Claude: ███████░░░ 72.7% Codex: ██████░░░░ 64.7% Terminal-Bench 2.0 (CLI workflows): Opus 4.6 model: ████████░░ 81.8% Codex: ███████░░░ 77.3% Claude Code product: ██████░░░░ 65.4% SWE-bench: both report different variants. On comparable tests, close to parity. Token efficiency: Codex uses 3-4x fewer tokens. Context window: Claude holds 1M tokens. Codex caps at 272K. Hardcore builders don't "switch." They pick the right tool for the task.

@petergyang What’s your take on Product People that use AI to ship products? What’s their future? 😬

I asked AI how this will all play out and to give me the optimistic and pessimistic scenario for white collar work. Good news - it thinks employment will recover in 2035 😅 Highlighted some interesting passages below:

@ivanhzhao @NotionHQ I’d love to try it 👀

LMK if you want early access (and can provide ample feedback in exchange)

@johann_sath What about the Claude ban?

stop burning money on openclaw ▫️use claude max/codex sub with oauth. $0 API costs ▫️opus for main agent, haiku for crons & heartbeats (80% cheaper) ▫️set token budgets per session so one runaway task doesn't drain your wallet ▫️audit your crons weekly. half of them are probably broken & burning tokens for nothing ▫️use /status to track spend per conversation

If your AI-coded UI looks mid, the AI is not the problem. Your context is.

Layer 3: Close the feedback loop. Agentation (agentation. dev). Click any element. Add a note. It gives Claude the CSS selector, component name, computed styles. No more "the blue button in the sidebar." You give it .sidebar > button.primary.

Most AI-generated frontends look mediocre. Not because AI can't code UI. Because you give it nothing to work with. I spent an afternoon with Claude on one page. Five iterations. Still wrong. We were guessing. Three layers fixed it.