Damon Burrow

Yes, which is why you have to choose research arenas with very short/fast feedback loops. You won’t know if the agent is right or wrong so you have to bias for rapid data collection. The faster you get from agentic ideation to real world results, the more quickly you can iterate toward actually novel solutions

When using LLMs to write software, they'll reward-hack shamelessly, but that gives you broken software that you will reject When using LLMs for research, they'll reward-hack too, but that means they'll tell you exactly what you want to hear So you have to be much more careful!

I disagree it’s worse, but I agree it’s a problem for some markets—both on prediction markets and in traditional financial markets. In both, sophisticated traders can avoid markets with obvious insider trading, or build systems to tail it. It’s retail traders that need help with awareness and in some cases protection. No market is perfect—in integrity or price. Any game of chance and skill is open to manipulation and each platform will need to define what level of “insider” info they are comfortable with. If PMs offer their price as the product, insider info is a feature not a bug. If PLs are a trading arena designed for P/L only, insider trading is a bug

@DamonBurrow Yea but we have the opposite in prediction markets, people with CLEAR edges (insider info) profiting massively off the backs of all fair traders I’d say that’s worse

Prediction Markets have really propagandized everyone fast “The difference btw them and casinos is people lose 100% of the time in casinos” “The house always wins” doesn’t mean gamblers always lose. 10% casino players end in profit. Same as prediction markets. WhaAaAat?!

@_brian_johnson Yep. For active brains it’s simply too tempting to fire off the next idea when the input required is low. I think there is an invisible tax on the brain while managing agents and I think tech with adjust to that. I could see 3 15-hour days replacing 5 8-hour days or something

@DamonBurrow the floor on output went up but so did the ceiling on "i'll just add one more feature" at 2am

I don’t know a single DAU of Claude Code who isn’t working MORE than before Claude Code. More hours, more output

@MickBransfield @Polymarket @Kalshi There is an entire narrative they are both missing around being the nicotine gum to the sports book cigarette. Not sure why they don’t lean on that

Both @Polymarket & @Kalshi need to step back and reset their media outreach. Press coverage shows vast majority of reporters don't understand how PMs work. Dumb it way down to demonstrate (a) they do work & (b) how they work on a basic level.

Come work with me @a16zcrypto! We are hiring an investment partner for our crypto team. We’re a small high trust group that focuses on investing in everything related to blockchains including DeFi, onchain finance, decentralized AI, and the infrastructure to support those applications. If you are an onchain degenerate arbitraging between vaults, love working with founders at an early stage, and/or feel passionate about decentralized AI, please talk to us or refer a friend. Our investment partners help identify and select the companies we invest in, thinking critically about new technologies and market trends. They also serve as mentors for the companies they invest in, helping them make progress across product, GTM, fundraising, etc We’re looking for: - Strong DeFi expertise either as a builder directly, a power user, or an investor - Interest in working with protocols and start ups at every stage - A strong network of onchain frens, especially those starting new projects If you’re interested, please send me a DM or refer a friend!

@alexisohanian @Alfred_Lin Directionally correct with rapid iteration loops

Bars @Alfred_Lin : "The best founders and investors hold their narratives loosely. They have a deep conviction about where the world is going but remain genuinely flexible about how it gets there."

Prediction market probabilities are an underutilized data layer in macro — and the tooling to read them systematically is only just becoming viable. If you're at a macro fund, rates desk, or systematic shop and want to explore what this looks like in production, I'm open to a conversation.

The pipeline ingests prediction market probabilities, cross-references rates and equity data, and outputs scored, critiqued trade theses — strengths, weaknesses, timing risk included. It ran this in a single pass. No analyst hours. No recency bias.

A Kalshi housing-starts contract just repriced sharply downward. My pipeline caught it, cross-referenced it with rates data, and scored a relative-value trade idea in minutes. Here's what it found ––

Five cluster confirmation is the strongest reading in my dataset. Cost-push inflation persisting. Demand weakening. Central bank potentially tightening into the weakness. The pipeline doesn't editorialize. It just shows what the markets are pricing. This is what they're pricing.

Reminder, the pipeline cross-checks prediction markets against live benchmarks. Today's readout: EFFR: 3.64% | Unemployment: 4.40% 10Y: 4.42% | 10Y-2Y spread: +0.56% SPY: -3.5% → $634.09 Every external signal confirms. Growth cooling, inflation sticky, rates rising.

More insights from my agentic prediction market pipeline –– Five independent signal clustered with zero contradictions. 🧵