Last Week in Crypto - Oct 9
The experiment continues.
Welcome to Last Week in Crypto. This is a mostly automated and algorithmically generated rundown of the most significant headlines in crypto over the last week. You can read a bit about how it works here. Reflections on building at the end.
Last Week in Crypto:
• NYSE parent company ICE commits up to $2B cash investment in prediction platform Polymarket, valuing firm at roughly $8B pre-money (via Wall Street Journal) 🔥🔥🔥🔥🔥
• SoftBank-backed PayPay secures 40% equity position in Binance’s Japanese subsidiary to integrate crypto trading with cashless payments (via Bloomberg) 🔥🔥🔥🔥🔥
• CZ’s YZi Labs unveils $1B fund targeting BNB Chain developers across DeFi, AI, and Web3 infrastructure (via CoinDesk) 🔥🔥🔥🔥🔥
• Ethereum Foundation establishes dedicated Privacy Cluster with 47 researchers and engineers to embed privacy features at protocol level (via Cointelegraph) 🔥🔥🔥🔥
• Polymarket CEO reveals two previously undisclosed funding rounds: $55M at $350M valuation in 2024 and $150M at $1.2B valuation in 2025 (via The Block) 🔥🔥🔥🔥
• MetaMask rolls out in-wallet perpetual futures powered by Hyperliquid and plans exclusive Polymarket prediction market integration (via The Block) 🔥🔥🔥🔥
• Gemini establishes Australian subsidiary with AUSTRAC registration to offer localized AUD crypto trading (via Reuters) 🔥🔥🔥
• Bitcoin-focused life insurance provider Meanwhile secures $82M funding from Apollo and Bain Capital following earlier Series A (via Bloomberg) 🔥🔥🔥
• Chicago stablecoin startup Coinflow closes $25M Series A led by Pantera, supports payments in 170+ countries (via CoinDesk) 🔥🔥
• Financial services API provider Routefusion raises $26.7M Series A for embedded accounts and compliance tools (via Crunchbase News) 🔥🔥
• Analysis examines growing risks of buy-now-pay-later services like Klarna on consumer financial health (via New York Times) 🔥🔥
• European financial regulator ESMA outlines plan to centralize oversight of exchanges, crypto firms, and clearing houses (via Financial Times) 🔥🔥
• Revolut CEO and major shareholder Nik Storonsky changes official residency from UK to UAE (via Financial Times) 🔥🔥
• Fight Fight Fight LLC, issuer of $TRUMP token, pursues $200M-$1B raise for digital asset treasury company (via Bloomberg) 🔥
Some reflections on building:
Initial feedback is that the fire emojis are confusing to people. I thought it was obvious everything is graded on a 1 to 5 fire emoji scale indicating significance and thought it made it fun and whimsical. Feedback says otherwise1. I’m leaving it in for a few rounds to gather more data.
Users are telling me categories and specificity would be good. This makes sense but I’ll have to think about how to implement it.2
One user3 asked for a video of an AI reading the headlines every morning, possibly on TikTok4.
It is possible a daily cadence would be better than weekly.
For some reason the AI likes to cite sources that have a paywall but then link to a version that doesn’t have a paywall. I am going to have to figure out how to build around this. It’s sort of weird because I’m not entirely sure why it happens and it happens a bit randomly.
Relatedly, AI being deterministic means I can run this thing twice in a row and get different results. It is really fascinating to watch. This thing is relatively simple so it doesn’t matter too much but this might be an opportunity to build a simple Eval framework5 to further refine results.
A hard problem emerging: getting random fundraising results removed. Right now I manually delete these because it takes all of 60 seconds but I’d like to automate this part too. So far my attempts to automate this out have all failed and some attempts even remove legitimate results6.
Cue the Simpson’s Principal Skinner meme…maybe the users are wrong?
There also becomes a problem where if you go down into a bunch of sub-categories of crypto, then there isn’t enough happening on a weekly basis and you end up with only like 2-3 headlines per sub-category and 1 of those is trivial.
This user was my friend who told me this while we were sitting in a Sauna and upon reflection I’m not sure if he was being serious or not.
If I have to use TikTok to make this successful I think I’d rather just see it fail.
If you follow the Expert AI PM zeitgeist then you know evals are all the rage. I think people are just hyping it up in the name of influencing. It’s just a test framework. You should always test things. No different than it has ever been with software.
The human job of reviewing results like this is safe for now!