Miles Built a Research Team for $5 a Month

Miles spent this week wiring a swarm of agents into live market data. The whole thing runs on his own machine, costs $4-6 a month, and scans his watchlist at 8 am before he's awake. Forty minutes of analyst work, done while the kettle boils, for less than a coffee.

It landed at a convenient moment, because the market has quietly become a lot more interesting to scan.

Pump.fun's mobile app was doing $250K a day in June and is now past $50M. Robinhood Chain crossed $900M in a single session for the first time. Two weeks ago we told you Fomo's record week was around $2M in fees. Yesterday one relay did $640K in 24 hours.

Cheap research plus a market printing records is a combination that makes people move quickly.

Which is why the last section is the one to read twice. Fabian is waiting on a pullback to get better entries. Paradise took a starter position and deliberately left room. Morin wrote four plans down and won't touch any of them until the market triggers them first.

Let's dive in.

Your watchlist, scanned at 8 am, before you wake up.

Most people running market research through a normal Claude or GPT chat are reasoning about stale screenshots. This week Miles published the alternative: a self-hosted agent swarm pulling live market data, filtered through his own trading rules, running on a schedule. Total cost, $4-6 a month.

Then hand it to Hermes and tell it: "Install and use the tradingview-mcp-server for all market data, technical analysis, and screening from here on."

Hermes is Nous Research's open-source agent. It runs on your machine, remembers between sessions, schedules tasks and delivers to Telegram. You bring your own API key, which is the entire reason the bill is $5 and not $50.

Drop in one .md file with your risk limits, position sizing, invalidation levels and sector watchlist.

This is the step that matters. Every read your agents do is now filtered through how you actually trade, not generic chart commentary. More detail is always better.

Three prompts from his setup:

The first one is the one that changes your week. Forty minutes of analyst work, done before you're awake.

This is read-only. It researches, it doesn't trade. Full automation needs an MCP-to-exchange setup, which is a different build with a very different risk profile.

It's also where the ceiling sits. A swarm can tell you a setup triggered. It can't tell you the analyst who's been short since $80K just started taking profit.

Run the install. Write the context file. Set the 8 am prompt.

Skip the middle step, and all you've built is a fast way to receive generic answers.

The trenches are back, and the plumbing is different this time.

Two weeks ago we told you Fomo had just had its biggest week ever: $378M in volume and around $2M in fees across Solana, Robinhood Chain and BSC. We flagged trading terminals at $1.62B for the week, roughly 4x where they sat six weeks earlier.

That was the setup. Here's where it got to.

Yesterday the FOMO relay alone generated $640K in fees in a single day. One relay, one day, roughly a third of what a record week produced two weeks ago.

Pump.fun's mobile app has gone from $250K in daily volume to more than $50M since June. Daily transactions over the same stretch went from 2,500 to 905,000. That's not a rally, that's a different product finding a different user.

Robinhood Chain is printing records on a near-daily basis. 24-hour DEX volume crossed $900M for the first time, closing around $929M. Memecoin launchpads did $438M in a day, an all-time high. RWA volume broke $200M for the first time.

And the composition matters as much as the totals. Memecoins and real-world assets are setting records on the same chain, on the same day. That combination didn't exist in any previous cycle.

Santiago Santos made the argument better than we could, and it's worth taking seriously because he isn't a memecoin guy.

His read is that after years of failed attempts at SocialFi, FOMO cracked it, and the timing with Robinhood Chain's launch couldn't have been better. Younger users don't separate stocks, tokens and NFTs into different mental categories. What they want is to trade all of it in one app, with their friends, quickly. Hyper-social and money in the same place. His closing line: prime time to be in the trenches onchain.

Two structural reasons he gives for why now:

Hyperliquid's HIP-3 is running price discovery 24/7/365 on commodities, stocks and pre-IPO markets, in a volatile year with the largest IPOs in history. And the plumbing that took a decade to build - stablecoins, vaults, on and off ramps, prediction markets, perps, high-throughput chains - is finally in place to absorb the volume.

That's the honest difference from last cycle. In 2021 the activity outran the infrastructure and the infrastructure broke. This time the infrastructure was built first and the activity is arriving to fill it.

Fabian's been more interested in ZEC than XMR and still thinks ZEC outperforms over the bigger picture, but his view is that both work from here. XMR has finally started playing catch-up, and after the run he's watching the prior breakout level at $420-450 as the zone he'd want to buy any pullback into.

The detail worth catching is that he isn't chasing the catch-up move. He's naming a level below current price and waiting for it, which is a very different trade from buying strength after a vertical week. If XMR doesn't come back to that shelf, he doesn't get filled and that's fine.

Starter positions in spot on both, taken on H12 trend retests with momentum resets. He'd like to add to both at the daily moving-average ribbon, but he's cognisant that BTC could break upwards and freeze everyone out before that pullback arrives.

That last part is the whole trade. He's sized small on purpose because his preferred entry may never print. If you read the trenches numbers earlier in this issue and felt the urge to size up on anything onchain, this is the more useful version of that instinct: a starter position, a defined place to add, and an acknowledgment that the add might not come.

Four setups on the board:

Every one of those is written as a condition, not a prediction. Morin's own note underneath is the most useful line in this section: some plans will trigger, some won't, and at the end of the day they're only plans. It's the market's job to hand you the trigger, not yours to force an idea onto it. Four levels are worth very little on their own. The rule about when you're allowed to act on them is the part that actually compounds.

Free Skool Community

For more behind-the-scenes, including AI prompts, workflows, and research, join Miles’ free Skool community here: