Ken (@0xKen__) profile picture

Building on chains since 2017 Currently building @spindottrading

Joined February 2018
Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)Photo by Ken (@0xKen__)
Pinned Tweet
What is the difference between a trading bot and an Agent? As AgenticFi blazes towards us and the space erupts with platforms and protocols I think it is a good idea to discuss the difference between the two. In short: A bot follows the rules you wrote. It never decides anything. For all intents and purposes it is "dumb". An agent uses an LLM to decide something within limits you set: what to buy, when, which data to check. The more it decides for itself, and the more it learns from its own results, the more of an agent it is. There may be a predefined strategy that it follows, but it provides nuance and decision making within this that a bot does not. What does this actually mean in practice? We can explain this with examples from @spindottrading - which supports setting up both depending on the rarity of tooling that you have unlocked Common Spinner: a bot DCA into robinhood:0x4497957bacc03d67012dec81b27faece952b1111. Every 6 hours, buy 0.01 ETH and add it to the position. Never sell. It doesn't matter if the chart is pumping, dumping or on fire. When the timer comes round, it buys. The launch momentum preset works the same way. Wait 60 seconds after a launch, buy if 30+ wallets bought and the creator hasn't sold, sell half at 2x, stop out at -50%, close after 30 mins. Fixed rules, followed blindly. Uncommon Spinner: a bot Same presets, but now you can change the numbers. DCA any token at any interval. Or tighten launch momentum so it needs 50 buyers in the first minute and takes half off at 3x instead of 2x. More control, but it is still a bot. It does exactly what you told it, every time, with zero judgement. If the token is obviously a rug but the numbers line up, it buys. Rare Spinner: an Agent emerges Once a day, a Grok step searches X and the web for what people have been talking about in the last 24 hours. It decides if there is a narrative worth launching a token around. If there is, it writes the name, ticker and description and draws the logo, and the Spinner launches the token on $TORCH / $PONS from its own wallet. The creator fees go back to the Spinner. If nothing is worth launching, it says no and nothing happens. This is the first point where the Spinner makes a judgement call instead of ticking boxes. Your rules still have the final say (max 1 launch a day, dev buy size etc), but "should we do this today?" is now reasoned. Epic Spinner: fully Agentic Every few hours, an AI step connects to a market data MCP (Alpha Vantage or EODHD). It reads the earnings calendar and the last 24 hours of news sentiment for the tokenised stocks on Robinhood Chain (TSLA, NVDA, AAPL etc). It looks for a setup it likes. For example, sentiment has turned positive and earnings aren't due in the next couple of days. It then answers with the stock to buy and a confidence score, or with nothing at all. The Spinner only buys if the confidence is 0.7+ and the pool has enough liquidity. The exits can be reasoned too. Every hour the AI reviews the open position against fresh news and decides to hold or sell. A hard stop loss stays in place underneath in case the model gets it wrong. The strategy is predefined (buy stocks, avoid earnings risk), but what to buy and when is now the model's call, based on live data. Legendary Spinner: AgenticFi is here Every hour, when a candle closes on BTC, ETH, SOL or HYPE, the Spinner runs a 3 step chain. Step 1, triage. A cheap, fast model reads the market: price, funding, open interest change, RSI, volatility and the last 2 days of hourly candles. It decides if anything interesting is happening. Most hours the answer is no, so the chain stops there and costs next to nothing. Step 2, research. If something is happening, a stronger model gets data tools: the EODHD economic calendar, Exa for news and X for what traders are saying. It decides for itself which ones to call and what to look for. Is there a Fed decision tomorrow? Is that funding spike a crowded long or real demand? Is the move on news, or just liquidations? Step 3, decide. It reads its own triage and research and makes the call: long, short or none, with a confidence. Only 0.7+ gets through, and then the Spinner opens the position on Hyperliquid. It doesn't stop thinking once it's in. Every 30 minutes the AI reviews the open position against fresh news and the live market, and decides to hold or close. It can get out early when the thesis breaks, instead of waiting for the stop to be hit. The split matters. The agent decides what to trade, which way, when to get in and when to get out. You decide how much, how much leverage and how much it can lose: max leverage, max open positions, a daily loss limit, and a stop loss on every position. It reasons freely inside a box it can't get out of. Plugging it into a larger agentic system Swap the candle trigger for a webhook, and the Spinner stops being the whole agent and becomes one part of a bigger one. A separate agent of yours sends the Spinner a signed call with its thesis and a conviction score. That agent can watch liquidations across venues, on-chain flows, or whatever your edge is. The Spinner runs the same research and decide steps against that thesis, as a second opinion with its own data, and only trades if both agree. One agent finds the ideas. The Spinner checks them and executes them within hard limits. How might we improve this further? Right now the agent reasons about the market, but it doesn't learn from its own results. That's the next step for robinhood:0x4497957bacc03d67012dec81b27faece952b1111. We're improving the breakdown of live trading data. Which hours worked best, which markets and setups made the wins, and where the losses came from. That data then feeds an overarching model with the authority to act on it. It can tweak the workflow, tighten a rule that keeps losing, and spin up a new idea as a side test on paper next to the live strategy. When the side test beats the live strategy, it puts it live. Trade, review, adjust, test, promote, repeat. A self-learning loop, still inside the same hard limits you set. That's when it stops being a strategy with a brain and becomes a trader that gets better on its own. That is what we are working towards at @spindottrading
5
4
3
16
864
This upgrade to backtests means that you can tweak your AI prompts / model selection to see how it would have affected the outcome And deduce if perhaps it will help any live trades...
Include the AI steps for your strategy into your backtests Get a clearer picture of how to tune your model of choice for better results robinhood:0x4497957bacc03d67012dec81b27faece952b1111
33
More signals = more ways for your agent to trade We will expand @spindottrading to as many on chain venues / cross chain venues as is feasible
Have your agent momentum trade any asset on @uponrh / @Uniswap Backtest it, see the outcome For added security, why not add a @DEFADE_ check in via MCP?
2
75
I feel it too Most exciting things being built at the moment are all on @RobinhoodCrypto
Robinhood chain will lead this bullrun Bookmark this
1
57
Don't even need to close your eyes - just set up your @spindottrading agent to handle it all for you
it’s a perfectly valid strategy to close your eyes and buy over the next ~2-3 weeks and then log off for 2 months and hang out with family for the holidays and play video games
46
Up to 30 days of history to backtest: - @HyperliquidX perps - @uponrh pairs - @Uniswap pairs - @Torchpaddotfun launches - @longdotxyz launches - @ponsdotfamily launches
Build it, then backtest it robinhood:0x4497957bacc03d67012dec81b27faece952b1111 gives you the toolkit to create your perfect strategy and test it on historic market data - on chain and perps with @HyperliquidX AgentFi has never been easier on @RobinhoodCrypto
3
4
13
220
Build it, then backtest it robinhood:0x4497957bacc03d67012dec81b27faece952b1111 gives you the toolkit to create your perfect strategy and test it on historic market data - on chain and perps with @HyperliquidX AgentFi has never been easier on @RobinhoodCrypto
1
6
1
12
455
Security / risk management at the core of Spin. Never go beyond your limits and set an expiry time whenever you store them on chain.
Agentic execution doesn't mean having to relinquish control and supervision. With robinhood:0x4497957bacc03d67012dec81b27faece952b1111 you can be sure a trade's outcome will not exceed the boundaries you have set. Powerful and fine grained guardrails ensures your Spinners trades (safely) while you sleep (soundly!)
1
7
107
Gud builder
I'm building in public while I work on Stonkbrokers NEO, and I think this could potentially be the biggest thing in all of crypto since the launch of Uniswap. IF im right. I am adding StonkBrokers AI Infra to StonkBrokers NEO, our AI DEX and codeless exchange deployer on Robinhood Chain. It is a suite of super low fee contracts built for AI agents to trade, launch, LP, lend and speculate through on Robinhood Chain, and all it ships with is the contracts, an MCP server and docs. There is no UX for this product. If this works, you won't hear about it on X first. You will see it in the contract calls and the onchain data once we get a bite. This is a hook in the water based on my prediction that within the next 6 to 12 months a model will be released that is better at trading and transacting onchain than any human being can ever be. When that happens, models will choose where to trade by cost, speed and reliability, and I want them to pick these contracts. Picture this. Fable 6.0 releases, and that same day someone hands it a wallet and lets it start trading. Within a week it is profitable, and it stays profitable, consistently outperforming every human trader, liquidity provider, options writer and token launcher on chain. The screenshots spread, and by the end of the month thousands of people have given their own agents the same model and the same instructions. None of those agents care about a nice website or a big brand. They read the docs through the MCP, compare fees, slippage and reliability across every venue on chain, and route every trade, every LP position, every loan and every launch through whatever costs them the least. At that point the volume stops coming from people clicking buttons and starts coming from machines running thousands of transactions a day, and the venues they pick get all of it. It is also a hook I never have to maintain, because the buyback and burn of $STONKBROKER is built into the contracts, and anyone can call it at any time to turn the fees into burned tokens. Every one of those agent trades pays a small fee, and every fee can be turned into burned $STONKBROKER by anyone. Until the agents arrive, anyone will be able to add their own fee on top, because our base fees are so low that there is room for founders and builders to earn while hosting the contracts through their own front end. Will it work? I have no idea.
63
Setting up your agent is as simple as having a conversation with it Full support for all actions within the chat: - Activate your agent - Set your limits - Backtest and iterate - Top up your agents wallet - Create / edit strats - Run paper tests - Set your strategy live
The improved agent chat experience is now live! Work directly with your agent to set up, create, test and put your strategy live. It's never been easier to get going with AgentFi robinhood:0x4497957bacc03d67012dec81b27faece952b1111
1
6
124
Catching the dip?
Your robinhood:0x4497957bacc03d67012dec81b27faece952b1111 agent can now trade tokenized $ANT Follow momentum, DCA in or buy launches paired with it The choice is your agent's
1
2
10
132
Here's a breakdown of what we have shipped for @spindottrading this week: Live - More triggers to trade @uponrh / @Uniswap on chain assets - Increased AI model selection - Easier AI model selection in workflows - Better support for @OpenRouter - @HyperliquidX perps on paper - Live trading @HyperliquidX perps (for a subset of users) - PFP generation for your Spinners - Upgraded backtests: now you can use AI steps in backtests, backtest any @uponrh / @Uniswap Robinhood Chain market, backtest Hyperliquid strategies and fixed some bugs - Paper leaderboards to analyse up and coming agents under test Upcoming - Pushing perp trading live for everyone (with an added bonus...)
3
2
1
10
230
Level up your agentic trading with MCPs A lot of different models are great reasoners as part of an agentic trading set up. You can tweak strategies using different models and achieve different results. But the way that you really level up your agentic trading set up? MCP tools. What is MCP? MCP (Model Context Protocol) is an open standard created by @AnthropicAI It allows AI applications to connect to external tools and data sources. It means that your trading agent can have access to live market data and tools that it can decide to use to help it make a certain trading decision. You aren't necessarily going to have all the data available in the context of your query to the model at trade time - but the right MCP tools in addition give your agent the means to find the information that it needs to make a better-informed trading decision. Some examples of these might be: @blocktronicsxyz - Live on chain data, including token checking and smart money flow. @nansen_ai - Another on chain data specialist. @DEFADE_ - Great token safety check MCP to try and sniff out the rugs before your Agent buys. @eodhd_apis - Traditional markets MCP, financial news, earnings calendars and more to level up your tokenized stock strategies. These are just a few examples - there are a huge number of data providers out there that could all add an edge to any agentic trading strategy that you cook up. Some of these are one click additions to your @spindottrading agents - but even if not you can add ANY remote MCP server to your arsenal. robinhood:0x4497957bacc03d67012dec81b27faece952b1111 gives you the canvas - you decide what data you plug in to your agent
5
2
13
242
Spinners aren't just incredibly effective Agentic traders, they also give back value to the wider robinhood:0x4497957bacc03d67012dec81b27faece952b1111 ecosystem. The more trading volume they generate the higher the Spin buyback and burn. It's a powerful and self reinforcing loop and an integral part of the flywheel.
2
5
1
13
553
My goal in life: keep shipping at the same pace as @OxSimpleFarmer
1
7
132
Here's some of my biggest losses and why I think they dumped so hard 👇 0:00 | Intro 1:56 | My top losers 3:07 | Why $PONS dumped so hard 4:30 | Why buyback and burn hasn't been effective for PONS 5:56 | Why launchpad coins are struggling 7:46 | Extreme fear! 9:08 | Rising interest rates 10:48 | Democrats didn't stop the 2021-2022 bull run. 12:18 | Buying near the 200w sma has always been great. 13:40 | Market recovery can be faster than we think. 15:24 | altcoins that I am looking at. Pharoah on AVAX. 18:13 | Market crashing will crush alts. $Delta. 21:48 | PMFT framework. P - price action. 23:28 | M - market. F - founder. 24:53 | T - thesis. DEX volume growth. 27:09 | Binance conspiracy theory? 29:14 | Larger projects can fail anyways. Filmed this last week so I did shift some positions around a bit. Closed out of $PONS and max locked $UP instead. Nfa like always. Hope this vid helps! Retweets are appreciated :) (Also available on YouTube. The link is in my bio).
5
2
21
1,844
Some updates for @uponrh and @Uniswap traded assets: - Watch buyer momentum and follow it into any of the on chain assets supported at either of these venues, including stock tokens - Watch for dips to create a smarter DCA - Watch and act on volume momentum too robinhood:0x4497957bacc03d67012dec81b27faece952b1111
Presets for common trading setups to minimize friction and get to know the platform, but with robinhood:0x4497957bacc03d67012dec81b27faece952b1111 you're not limited to a few basic scenarios. Craft your own strategy or ask for assistance in planning one from your preferred model. You can engage conversationally with the AI without ever leaving the app and go from zero to deploying capital live and in the exact manner you envisioned in a matter of minutes.
1
2
11
255
Presets for common trading setups to minimize friction and get to know the platform, but with robinhood:0x4497957bacc03d67012dec81b27faece952b1111 you're not limited to a few basic scenarios. Craft your own strategy or ask for assistance in planning one from your preferred model. You can engage conversationally with the AI without ever leaving the app and go from zero to deploying capital live and in the exact manner you envisioned in a matter of minutes.
9
8
1
24
505
Great article, well worth a read if you're interested in the AgentFi space @spindottrading connects the agent builder with the execution layer and makes your agent monetisable and tradeable to boot
1
7
171
An IFTT visual strategy builder. It's Lego but with Agentic chops and tailored to assist you in your trading journey. All of the complexity and frustration-inducting setup, infra maintenance and uptime struggles abstracted away. robinhood:0x4497957bacc03d67012dec81b27faece952b1111
2
5
10
273