Trade six types of strategies. If you have momentum trades, add in a system for mean-reversion. If you have trend-following trades, add some carry strategies. 5 from one strategy, could be -4 from the other and it'll auto-reduce scale. h/t discussion w/ @SimpelAlpha. Thread👇

You get this options/futures dynamic because of:. 1. Capital controls.2. Asians love to gamble. So India, South Korea definitely. Not really: Taiwan. Asked my friends why and they said no capital gains tax, so they just do stocks. The answer seems too simplistic for me.

Did a drive-yank on the server and it took longer than I thought. I think I need to remove this step from the fire-checklist, lest I get caught in the smoke.

Set all your clocks to UTC, and you can be like me: on time for things happening globally, but late for everything local.

Here is a related thread on building the context.

In the meantime, I encourage the power users to jump around the tools while the VC-subsidized era lasts and only bring in the tools that improve your context scaffolding, and throw away everything else. 10/.

There's a couple of things Cursor could do to change course, but it'll take more VC money to pull it off. Not including it here because this thread is long enough (it's related to upskilling the normal user toward power users w/ more tuned skill-tree rails). 9/.

Anthropic is much better at context scaffolding with Opus for planning, and optionally Sonnet for Action. They currently can't beat Gemini's context window, and this is likely where the next push is going. 8/.

Context scaffolding is getting the best context into the system instead of embedding-search-request context. 7/.

Cursor will fail because their journey to grind down unit economics captures more of the "normal user" segment. They gives up the "power user" demand and miss the next wave of context scaffolding.6/.

This is why Claude Code is winning, because Anthropic made optimizes for context and NOT time-of-use:.- Cursor's time-of-use model uses embeddings to find context needed (good for cost!).- Claude Code's context model will scan files to grab the full context (great for coding).5/.

This is a mistake, because normal users don't complain about the tool becoming worse since they don't notice. The complaints you hear are from power users, an unprofitable segment overall (demanding VC-subsidized models). 4/.

Cursor is continuing to target the normal user because:.1. There's a lot of them.2. They've achieved product market fit.3. They can grind down the unit economics to (hopefully) become profitable. 3/.

Companies roll out Cursor because it turns "normal users" into "stronger devs," not because it makes them "power users." . The power user workflow is heavily focused on context scaffolding instead, so Cursor is not the right tool. 2/.

Although I rolled out Cursor, I'm replacing it with Claude Code. The Cursor pricing drama is happening because Cursor is unprofitable and will fail. because their business model is forced to chase cost-optimizations instead of becoming the best at context scaffolding . 1/

Didn't notice because of the switch to Claude Code. but there's gotta be a Claude Code pricing update soon, because only power users will get the max plan. and power users are exponentially more expensive to Anthropic. It's adverse selection, always.

I built my own agent stack to avoid the “ambiguous model problem. When you offload the feature to a company, they may switch to a weaker model (sonnet instead of opus) to save money . In the prompt I have a tag like “override_model:o3” and the backend will force a model/provider.

This makes sense to me, I don’t even pick up the phone unless it’s Pagerduty.

Claude overloads often, yes. At this point, I think it's a marketing tactic. A good product is one people want, a great product is one people can't live without.

I picked up one of these mini PCs because I wanted a firewall appliance with a 2.5Gbps NIC, it’s overkill but was the easiest way to get all the specs I wanted.

To use gemini with your auth inside a sandbox, do this:. 1. Install gemini and launch in a non-headless env . npm install -g @google/gemini-cli.gemini. 2. Auth with your google login. 3. Mount your ~/.gemini folder into the sandbox. services:. example_sandbox:. volumes: