AI Dispatch #38: Can AI Really Trade?
AI Dispatch #38: Can AI Really Trade? — research from Castle Labs.
Welcome to another edition of the AI dispatch, which starts with a question:
How good are AIs really at trading?
Last week, @jay_azhang of https://x.com/the_nof1 launched a competition named “Alpha Arena”, where they gave $10k each to the six most well-known AI models and asked them to trade.
How are they doing? Here is the last update:
Kinda funny how this was described as “AI hamster races” for those who remembers lol, but regardless, this may prove a worthwhile experiment to see which model performs the best in a trading scenario.
An interesting thing to note is that the different models have their own style of trading, as noted by @jinglingcookies below.
David from USDai shares more on the link between AI and stablecoins and how the latter can help it scale.
USDai is doing extremely well. We talked briefly about them in our Arbitrum corner this week: they have over $500m in TVL, and every time they open their vaults, they fill it.
Given the demand, the token has actually been trading with a premium above $1 for a while (probably will continue until caps are raised).
In other news, Mode launched Season 6, powered also by @optimism rewards, and focusing on AI trading agents:
These can be either:
Breakout Agent: with a focus on volatility
Trend Agent: focusing on trend momentum
The most important aspect here is Mode’s focus on ensuring verifiability of trading operations with a (soon-to-be-launched) framework built on ERC-8004.
More on AI verifiability from Eigenlayer:
Plus, someone built a game using EigenAI x Eigen Compute!
Someone recently integrated $REI’s closed beta access into his ChromBot LMAO
Depending on where you stand, here are some developments that you either consider extremely creepy or just the way our future looks.
Uber is now paying drivers for doing tasks such as:
Data labelling for AI
Recording audio samples
Language training across different scenarios
Uploading restaurants menus
Great post by @bearlyai, which highlights how data labelling is expected to continue growing as a market. Uber’s move is not random, as it recently acquired Segments AI, a data-labelling startup.
We haven’t been talking about Virtuals in a while! This week, they shared an analysis on their Agentic Commerce Protocol (ACP) and how it compares with others, such as Anthropic, Coinbase, Stripe, Google, and Ethereum.
For more Virtuals news, we also suggest this great interview the chads at Blocmates did!
Also, a lot of news for Bittensor, which has been benefiting TAO, Sammy sums them up:
Last week we also had our own couple articles on AI.
First, we shared more about the Almanak agentic wars and how they are planning to leverage incentives in a way similar to Curve but adapted to their protocol.
Secondly, we wrote an introduction to Robotics - which you should expect to be followed by more.
Now, Kaito launched a leaderboard for Robotics, which might be a treasure trove in terms of who to follow to get the best robotics news.
Among the projects that also launched their own leaderboard, we find OpenMind (@openmind_agi), which “ builds open-source software that helps machines think, learn, and collaborate, with OM1 - the Android for robots, and FABRIC - the coordination layer for machine economy.”
Those interested in diving deeper into OpenMind should do so with @cot_research's great report on the project!
Maybe not directly AI related, but super nerdy read on chip printers!
This week, we leave you with a bullish outlook by Galaxy Digital Head of Research!
Actually not, we leave you with a final reflection. As the demand for AI grows in the coming years, this will require building many data centres all around the world, which need water and fuel to run 24/7.
How will this impact our consumption of energy?
Imagine a data centre being built right next to where you live, sucking up all the water and electricity. Kind of a big deal, no?
If this is not enough here’s more weekly AI content from Jeff the man himself:
Anyway, that’s it for this week! Have a great Monday, and make sure to keep up because these are the best times to be on top of the market.
Originally published in the Castle Labs newsletter. Subscribe at research.castlelabs.io.