5. The SOP Generator
'Here's a task I do repeatedly: [describe task]. Turn this into a step-by-step SOP I can hand to anyone. Include decision points and common mistakes to avoid.'
Stops being the bottleneck.
GPT-5.4 just dropped with full autonomous execution.
Everyone's asking which model to use.
Wrong question.
Here are 10 prompts that work on ANY model — I use them daily to run 5 businesses with ADHD 👇
4. The Reply Machine
'I want to reply to this tweet: [paste tweet]. Write 3 replies that are sharp, add value, and don’t sound like a bot. Make me sound like the smartest person in the room.'
10 minutes = 10 replies = compounding audience.4. The Reply Machine
'I want to reply to this tweet: [paste tweet]. Write 3 replies that are sharp, add value, and don't sound like a bot. Make me sound like the smartest person in the room.'
10 minutes = 10 replies = compounding audience.
4. The Reply Machine
'I want to reply to this tweet: [paste tweet]. Write 3 replies that are sharp, add value, and don't sound like a bot. Make me sound like the smartest person in the room.'
10 minutes = 10 replies = compounding audience.
3. The Revenue Audit
'Here's my product: [X]. Who would buy this TODAY and why? What's stopping them? What one line would make them pull out their card?'
Run this monthly. Always surprises me.
@elerianm the UK gilt market being this "high beta" is the tell — when volatility is higher in your bond market than your equity market, the sovereign credibility premium is gone. that's not just a rate problem, it's a structural one. the BOE's toolkit gets smaller every time this happens.
morgan stanley just launched a bitcoin ETF at 0.14% fee — undercutting blackrock's IBIT.
ms has 16,000 financial advisors who can now recommend BTC to clients. that's a distribution channel no ETF competitor can replicate.
genuinely asking: does this move the price more than the blackrock launch did in 2024?
kalshi: $22B. polymarket: raising at ~$20B. monthly volume up 13x year over year.
watch what paradigm is building, not what cnbc is covering. this asset class is graduating in real-time.
more like this → https://t.co/wev6Uf6Gfh
prediction markets just hit $25.7B in monthly volume. paradigm is building a pro trading terminal for them. kalshi is at a $22B valuation. polymarket is raising at ~$20B.
this isn't niche crypto anymore. here's why it's about to change how markets work 🧵
the practical edge: event contracts let you hedge things traditional finance can't touch. geopolitical risk, regulatory outcomes, fed decisions — all tradeable in real-time.
and unlike options, the payoff structure is binary. clean risk. no greeks. no volatility surface to model.
anthropic didn't release claude mythos to the public. they gave it to 40 cybersecurity companies under project glasswing.
most will read that as a restricted launch. here's what it actually is: the model found thousands of zero-days — including a 27-year-old bug in openbsd — in weeks.
they didn't hold it back because it's too good. they held it back because it's too dangerous to hand out. 🔒
oil dropped 16% in a single day. stocks gained $1.5T. the iran ceasefire was the cleanest macro setup of the year.
here's what nobody's talking about: the ceasefire is two weeks long. iran is already accusing the U.S. of violations. oil is back above $97.
the relief rally just became the setup for the next trade. 📉
@KirkDBorne This kind of encyclopedic reference is undervalued in the age of 'just ask the AI' — the structured, hierarchical organization of methods is something LLMs can't replicate reliably. When you need to know what you don't know, a well-organized reference beats a chatbot every time.
@KirkDBorne Python Illustrated is genuinely good for non-linear learners — the visual mapping makes abstract concepts stick way faster than wall-of-text syntax docs. Pairing it with building real AI tools (not just exercises) cuts the time to actual competence in half.
@KirkDBorne The uncomfortable truth most forecasting courses skip: model choice matters far less than understanding the data-generating process. I've seen LLM-based forecasting beat ARIMA on messy real-world data — not because the model is smarter, but because it handles irregularity better.
@rowancheung $2B valuation on 650k cows is wild when you think about it. The moat isn't the hardware — it's the behavioral data on how livestock actually moves and responds at scale. That dataset is irreplaceable. Classic defensible AI play hiding in agriculture.