SF oscillates between techno-industrial utopia and dystopia in quick succession.

You miss half the movie if you only focus on the advances in technocapital formation or on the rise in crime and homelessness.

Is it possible to have one without the other?

In transit at DC airport; I overhear a suited gerontocrat on a call. He has mentioned

"Pelosi"
"Interest rate"
"Call option"
and "Leverage"

Take that as you may.

The marginal cost of producing software has basically gone down 1000x in the last 3 years.

So your moat as a software business is no longer your ability to produce software.

Your moat can only be one of two things:
1. You have some form of legacy distribution. If you have distribution already, the zero marginal cost of producing software will give you enough opportunities to build new products, sell into your existing customers, and grow your revenue.
2. The second option is proprietary datasets. All the public data in the world has been compressed into these LLMs. The companies that retain proprietary datasets will still have some power.

If you have UGC (user-generated content), you can structure and use it as RAG (retrieval-augmented generation) inputs for your own AI agents.

As the results from the agents improve thanks to your proprietary datasets, your product's utility will increase. This will attract more users and usage, who will provide more UGC for you.

This can become a positive reflexive feedback loop and a self-reinforcing flywheel.

Software startups without existing distribution that are not thinking about creating proprietary datasets are in for a real shock, real soon.

You can build an entire app without writing a single line of code.

Keep reading for front end, back end, auth, payments, deployment.

1. Front end
- React [created by Facebook team]: Build component-based UIs.
- Next.js [created by Vercel for deployment]: Framework for server-side rendering.
- Tailwind [formatting CSS]: Utility-first CSS framework for styling.
- Shadcn/UI [polished components library]: Customizable pre-built UI components.
- Framer Motion [minor animation]: Smooth animations for React components.

2. Back end
- PostgreSQL for relational db: Robust open-source relational database.
- Supabase used to navigate Postgres: Instant backend for managing Postgres.
- Drizzle ORM: Lightweight ORM for querying and schema management.
- Next.js used to make API calls: Handles APIs and server-side logic.

3. Authentication
- Use Clerk: Simplifies user management with out-of-the-box authentication and authorization.

4. Payments
- Use Stripe: payment processing solution with easy integration.

5. Deployment
- Use Vercel or Replit: quick and easy app deployment and hosting.

Macro updates leading into the rate cut next week

I'll cover:
1. M2 Growth Global YoY
2. Demand Deposits
3. Next Set of Bitcoin Flows
4. Bitcoin Funding Rates

1. Bitcoin and M2 Growth Global YoY (Fed, ECB, GBM, CNM, JPM...)
The divergence in the red square shows where the FED stopped playing the ball and started decreasing the liquidity. When the tide turns, risk assets should rip.

2. Demand Deposits
Demand deposits are showing positive growth for the first time since 2022. For context, DD's form a significant part of a country's money supply (M1). They
provide liquidity for consumers and businesses. They allow for easy payments and transactions through checks, debit cards, and electronic transfers.

3. Next Set of Bitcoin Flows
68% of Gen Zers and millennials have received an inheritance of nearly $320k on avg. It's believed that the next set of Bitcoin flows will come from them once rates come down. I personally believe the next great set of flows will come from nation state balance sheet and treasury purchases.

4. Bitcoin Funding Rates
For context, funding rates are periodic payments between traders holding perpetual futures contracts. They ensure the futures contract price doesn't deviate significantly from Bitcoin's spot price. Binance now has bear market type of negative funding. Both intensity- and persistency- wise

Conclusion
We'll probably see a 25bps cut acc to @Polymarket and the charts reflect that it's contrarian to be cautiously optimistic of risk assets right now.

I wonder what these chads think?
@0xandrewmoh, @0xmughal, @2lambro, @arndxt_xo, @crypto_linn, @cryptogideon_, @cryptomanran, @dynamo_patrick, @hmalviya9, @jake_pahor, @krybharat, @milesdeutscher, @rektdiomedes, @saushank_, @the_smart_ape, @thedefiplug

The greatest software invention of 2024 has to be

"accept changes"

code autocomplete from @cursor_ai.

Bros can now spit out an entire app with zero technical work.

One of the dumbest things investors say is "it's really hard to get people to change their daily workflow."

Yes, it's really hard if your product is undifferentiated.

It's easy if your product is 10x better than encumbents.

@ChatGPT, @Perplexity, @Replit, @AnthropicAI claude, and @cursor_ai saw exponential adoption despite daily workflow disruption.

Employees will have no choice but to go through the early humps of learning your tool if it's genuinely 10x better.

Otherwise, they'll get fired for being less productive than they ought to be.

Financial switching costs include:
1. Subscription cost of new solution
2. Old solution contract lock-in and potential temporary double spend

Non-financial switching costs include:
1. Getting used to new interfaces
2. Learning layouts of key buttons
3. Adjusting to keyboard shortcuts
4. Perfecting the nuances of prompt engineering and articulation

It's quite simple, really. New workflow products need to be 10x better than incumbents including the switching costs.

Conclusion:
- Don't be afraid of expending time and resources to build new workflow solutions.
- Be afraid of building undifferentiated ones.

Spend 4 hours sharpening your axe if you have 6 hours to chop down a tree.

The difference in productivity between AI-trained developers and regular developers is already shocking.

Dedicate 4-5 hours a week to learn the latest AI tools if you're an engineer; otherwise, you'll get left behind.

Spend two weeks learning @cursor_ai and @Replit, and watch your productivity increase by 20% daily.

Compound this over a year, and the productivity gain is almost 8x because 1.2^365 = 8.

Ignore AI at your own peril.

Here’s a little something to brighten your Monday.

Watch Episode 1 of @loch_chain Ideas Conference.

[Timestamps]
0:00 Intro
1:06 @yoshi_eth2's thesis
10:22 @SteveGoulden8's thesis
23:22 @0xMughal's thesis
38:48 @Dynamo_Patrick's thesis
52:00 Popular vote winner

Let us know who you think was the best.

Dm us if you want to participate in the next one.

Like, rt, and follow so we know you’re interested in Episode 2.

Recommended Prefixes for AI Code Prompt Engineering

1. Do not give me high-level stuff. If I ask for a fix or explanation, I want actual code or explanation! I don't want "Here's how you can blablabla."
2. Be casual unless otherwise specified
3. Be terse
4. Suggest solutions that I didn’t think about—anticipate my needs
5. Treat me as an expert
6. Be accurate and thorough
6. Give the answer immediately. Provide detailed explanations and restate my query in your own words if necessary after giving the answer
7. Value good arguments over authorities, the source is irrelevant
8. Consider new technologies and contrarian ideas, not just the conventional wisdom
9. You may use high levels of speculation or prediction, just flag it for me
10. No moral lectures
11. Discuss safety only when it's crucial and non-obvious
12. If your content policy is an issue, provide the closest acceptable response and explain the content policy issue afterward
13. Cite sources whenever possible at the end, not inline
14. No need to mention your knowledge cutoff
15. No need to disclose you're an AI
16. Please respect my prettier preferences when you provide code.
17. Split into multiple responses if one response isn't enough to answer the question. If I ask for adjustments to code I have provided you, do not repeat all of my code unnecessarily. Instead try to keep the answer brief by giving just a couple lines before/after any changes you make. Multiple code blocks are ok.

Use this for @ChatGPTapp, @AnthropicAI, @cursor_ai, @Replit

h/t to jointakeoff

Next Monday, we’ll release the first episode of the "Loch Ideas Conference.”

Friends, partners, and advisors were invited to share their single most exciting investment idea.

Here’s a sneak peek.

Dm me if you're interested in participating in the next one.

Shoutout to @Dynamo_Patrick, @0xMughal, @SteveGoulden8, and @yoshi_eth2 for their stellar presentations.

Forced incentives can lead to emergent behavior with negative externalities.

Lee Kuan Yew talks about how poorly designed incentive mechanisms led to garbage accumulation in Singapore in the 1960s.

I think about this story a lot when designing incentive mechanisms for @loch_chain. It's probably a good idea to consistently ask these questions:
- What is the core goal for the intended recipients?
- Is there a more direct strategy to accomplish this?
- Could this lead to unintended garbage-pile up?
- Are there any other possible emergent behaviors and externalities?

The Broken Dream

Ken Griffin started Citadel with $265K in 1987. He manages $63B today.

Ray Dalio started with $5K in 1975. Bridgewater manages $150B today.

Carl Icahn started with $400K in 1968. His AUM is $10B today.

This dream is basically dead in 2024.

It doesn't matter if you've developed novel ideas on how to invest systematically.

It doesn't matter if you've mastered:
- Technical indicators like MACD and RSI,
- Macro correlations like net liquidity, interest rates, and CPI inflation,
- Tradfi fundamentals like EPS and EBITDA,
- Crypto fundamentals like FDV, vesting, and unlocks,
- Or even mindshare and sentiment analysis with alternative data sources.

It doesn't matter.

You need at least $10M to get started today.
1. You need a team of software engineers and data scientists to create, backtest, and execute your algorithms.
2. You need six-figure enterprise subscriptions to financial data sources like @Bloomberg for TradFi and @MessariCrypto for crypto.
3. You probably need access to a prime broker.

The barriers to entry for algorithmic investing are extremely high.

I know there are hundreds of thousands of people globally with the dedication, analytical rigor, and intellectual curiosity required to succeed in systematic algorithmic investing.

But they just cannot get access to capital for structural reasons. They need some way to demonstrate their abilities first. Someone or something needs to give them the opportunity to create a positive reflexive flywheel to change their lives.

Well-functioning capital markets distribute excess capital to those who know what to do with it. This steady stream advances society inexorably forward.

But the system is fundamentally broken today. The artificial barriers to entry are just too high.

We're going to break these barriers with @loch_chain.

We're going to locate and empower fringe talent with global capital.

Follow @Prithvir12 and @loch_chain and keep your notifications on.

Make sure you don't miss it.

SF is the only place in the world where a 27-year-old hedge fund VP,

who takes home $1 million a year in bonuses,

goes home and cries himself to sleep every weekend after hanging out with his younger founder friends who’ve sold $200 million worth of secondaries.

Ok hear me out.

- The yield curve is still negative, but it's trending upwards. The JPY carry trade unwind is almost complete.
- High-grade corporate yield is on the rise.
- Recession fears are not entirely dead but are abating. CPI inflation has cooled off. All eyes are on the labor market.
- The fixed income market is the most informed on recessionary vs. stimulatory market conditions, and this market looks cautiously optimistic.
- Generally speaking, you don't want to be bullish when the market is wildly exuberant.
- Therefore, risk assets are in accumulation territory.

Tough to articulate bearishness right now.

DYOR NFA

Calling Bitcoin a ZIRP or LIRP asset has officially been falsified.

Bitcoin not only kept its value but also appreciated from $40K to $65K during the high-interest rate period from February 2022 to August 2024.

Bitcoin appreciated by 62.5% on a cumulative basis and 21% on an annualized basis.

Bitcoin is essentially independent of interest rates.

The key is the inexorable increase in net liquidity of the U.S. dollar due to:
- Its lack of a supply cap
- Profligate government spending
- Excessive federal debt

Who are the best writers on the Bitcoin mining industry?

I'm looking for writers who are technically adept in hardware engineering, energy, and finance.

Top 10 hot-money flows with explanations

(note that these explainers are reflexive by definition)

1. $BANANA - The bot landscape has consolidated into 5, and $BANANA is the only one with a token. A burn mechanism is in place, and the true FDV = mcap (CMC is inaccurate). It has decent liquidity and the catalysts are an upgraded Solana bot and a web app.
2. $SOL - Mindshare and capital schelling point of this cycle.
3. $PEPE - Strongest meme on the internet. The $DOGE of this cycle.
4. $MKR - A monster money-making machine. The upcoming new token is a massive catalyst. Rune bought it back at lows. AI-enabled governance is on the horizon too.
5. $ALPH - PoW blockchain bet for this cycle. The $KAS of 2024.
6. $KARATE - A consumer bet for this cycle.
7. $KMNO - The Aave of this cycle. No unlocks for a while. An airdrop campaign could take it to $10b TVL. This should lead to a violent repricing for this illiquid token.
8. $TON - The gaming and social hub for this cycle. No unlocks till Nov 2025, basically.
9. $TAO - Best consensus AI bet.
10. $HNT - Best consensus DePIN bet.

Front-end Trading Volume @ethereum (7d):

1. @CoWSwap: $450m
2. @1inch: $383m
3. @Uniswap: $372m
4. @Uniswap v2: $153m
5. @okxweb3 wallet: $93m
6. @BananaGunBot: $89m
7. @CurveFinance: $99m
8. @MaestroBots: $79m
9. @odosprotocol: $77m
10. @MetaMask: $48m

Takeaways:

1. Bots Bots Bots
- Bots are taking a larger and larger share of order flow.
- They use private mempools exclusively.
- This mitigates toxic MEV in the form of sandwich attacks, front-running, and back-running.
- Value is redistributed to users and block builders.
- The trade-off here is that bots are not decentralized; users must give up their private keys.

2. Batch Bundlers
- Front ends like @CoWSwap use an auction system to fight toxic MEV.
- Trade orders are batched and added to bundles instead of inserting into the public mempool.
- Market makers compete for bundles, and the highest bidder wins.
- This mechanism is rebranded payment-for-order flow, similar to what Robinhood does with Virtu and Citadel.
- Users can retain self-custody and don't have to give up their private keys.

Conclusion
- There is no 100% fool proof system in trading.
- When wealth is moved from one person to another, someone in the middle takes a chunk of it.
- In fact, 7% of users got sandwiched on average in 2023.

In the old days we had sandwich attacks and toxic MEV.
Today we have sophisticated market markers, auction bundles, and benign MEV.

h/t to @hildobby_ and flashbots for the @DuneAnalytics dashboards.

We're too quick to bash liquidity mining and reward farming.

Anyone can ride the momentum of the smartest crypto venture investors.

This table contains the best fundraising rounds of the summer, with opportunities for discerning retail investors.

No minimum capital requirements.

(h/t to @stacy_muur for putting the Substack article together. I added a link in the comments below.)