Weekly report · Nov 16 – Nov 22, 2026 · verified from public data
AI trading bot report: week ending Nov 22, 2026
We added up what 1,393 users of five AI trading apps on Hyperliquid made in the 30 days to Nov 16. At four of the five apps, the typical user finished within 25 cents of zero.
Week ending Nov 22, 2026. Data as of Nov 20, 2026, from the same files as the leaderboard. Past results don't predict future results. Not financial advice.
The story this week
We checked what people using AI trading apps actually made. The typical user made nothing.
Every AI trading app shows you its wins. Hyperliquid shows you everything. Each order placed through an app carries that app's tag, in public, so we added them up.
Between Oct 18 and Nov 16, five apps that say they run AI agents for their users handled $381 million of trades for 1,393 users. The apps are tread.fi, DXAP, Senpi, TrueNorth and Wayfinder.
The typical user did not get rich and did not get wrecked. At four of the five apps, the user in the middle of the pack finished within 25 cents of zero. At tread.fi, the middle user was down $7.93.
About half of users closed the month in profit at four apps, between 48% and 56%. At tread.fi it was 19%, for a reason explained below.
The totals are the interesting part. Senpi's 676 users made $411 between them. Senpi collected $518 in fees on their orders. Across all five apps, users closed $92,278 in profit and the apps collected $30,211 in fees.
Users of five AI trading apps on Hyperliquid, Oct 18 to Nov 16, 2026
| App | Users | Closed in profit | The typical user | All users together | Fees the app collected |
|---|---|---|---|---|---|
| tread.fi | 262 | 19% | down $7.93 | up $63,839 | $27,886 |
| DXAP | 223 | 53% | up $0.24 | up $24,980 | $1,068 |
| Senpi | 676 | 48% | down $0.01 | up $411 | $518 |
| TrueNorth | 213 | 53% | up $0.10 | up $828 | $0 |
| Wayfinder | 19 | 56% | up $0.16 | up $2,219 | $739 |
This edition covers the five apps with the most users. The other four on the board each had 14 users or fewer as of Nov 19.
None of this says AI cannot trade. It says that so far, for the person in the middle, an AI trading app has been a way to end the month where they started.
What the numbers do not show
- Thirty days is short. One good or bad month says little about any app.
- Closed trades only. Positions still open on Nov 16 are not counted.
- Only orders that carried the app's tag. A position opened in an app and closed somewhere else is missed.
- tread.fi is a different kind of tool. It runs market-making and hedged strategies as well as AI agents. A user who hedges on another venue can show a loss here and be level overall. Its 19% must not be read as "81% lost money".
- Users, not the AI. Several of these apps also take manual and copy orders. The numbers show what the apps' users made, not what the AI decided.
- Before Hyperliquid's own trading fees. Also, a wallet that used two apps counts twice in the 1,393.
Right of reply
Strategy Lab this week
That day's board →Every open-source bot we test gets $1,000 of paper money and the same rules. Spot only, live prices.
17 of 181 bots were up on Nov 20, 2026, day 13 of the record. Holding the same 20 coins stood at $788.57, and 166 bots were ahead of that. Best: ClucHAnix_5m1 at $1,033.80. Worst: Magic_Trailing_Stoploss at $82.18. With so few days, most of the difference is luck.
Real money this week
Vaults, pools and agents trading real money, from public venue data. Returns are before the operator's profit share.
TAS Agent Index (beta)
How it's computed →The median 30-day return of the 15 vaults and pools (5 AI-labelled) that pass both gates, out of 28 we track: 0.0%, against −26.4% for BTC over the same 30 days. 15 of 15 beat BTC. Last report (Nov 8): 0.0% across 17, so the index moved −0.0 pts. Lighter data ends Nov 19: BTC −25.3% over those 30 days. Before the operator's profit share; not an investable index.
Top 5 by return this week
Full leaderboard →Among the vaults and pools that pass both gates (the venue scores 50%+ on our Safety score, and the account never fell more than 30% from a peak), ranked by return over each one's own last 7 days of data. Returns are before the operator's profit share.
| Name | 7 days | 30 days | Since start | Max drawdown | Safety gate | Action |
|---|---|---|---|---|---|---|
| 1 +convexity Vault · Hyperliquid · since Nov 26, 2025 | +8.1% Nov 13 – Nov 20 | +17.3% | +65.0% BTC −31.6% | −12.2% | Pass Hyperliquid 50% | How to deploy → |
| 2 Orbit Value Strategies Vault · Hyperliquid · since Dec 17, 2025 | +3.5% Nov 13 – Nov 20 | +12.0% | +4.7% BTC −28.2% | −19.8% | Pass Hyperliquid 50% | How to deploy → |
| 3 AceVault Hyper01 AI-labelled vault · Hyperliquid · since Nov 26, 2025 | +2.1% Nov 13 – Nov 20 | +4.6% | +21.6% BTC −31.8% | −21.1% | Pass Hyperliquid 50% | How to deploy → |
| 4 Akka Hyper AI AI-labelled vault · Hyperliquid · since Nov 26, 2025 | +1.8% Nov 13 – Nov 20 | 0.0% | +9.5% BTC −31.8% | −5.0% | Pass Hyperliquid 50% | How to deploy → |
| 5 FuturAI Labs - High Sharpe, Low Volatility AI-labelled vault · Hyperliquid · since Mar 4, 2026 | +0.8% Nov 13 – Nov 20 | +2.0% | +9.6% BTC −15.0% | −1.9% | Pass Hyperliquid 50% | How to deploy → |
Worst 5 by drawdown
The deepest falls from a peak among everything we track, gated or not. A fall of more than 30% fails our gate: we show it, and never rank it. If you run a bot yourself, the setup guide shows how to cap what it can lose.
| Name | Max drawdown | 7 days | Since start | Gate | Action |
|---|---|---|---|---|---|
| 1 Long LINK Short XRP Vault · Hyperliquid · since Nov 26, 2025 | −95.3% | −32.6% | −92.6% | Fails fell more than 30%: shown, never ranked | Set up safely on Hyperliquid → |
| 2 drkmttr Vault · Hyperliquid · since Dec 17, 2025 | −87.4% | −42.5% | −69.7% | Fails fell more than 30%: shown, never ranked | Set up safely on Hyperliquid → |
| 3 [ Systemic Strategies ] Peter Schiff's vault Public pool · Lighter · since Mar 6, 2026 | −80.8% | – | −72.0% | Fails fell more than 30%: shown, never ranked | Set up safely on Lighter → |
| 4 AILAB TEST ULTRA Vault · Hyperliquid · since Dec 31, 2025 | −73.5% | +25.0% | −19.6% | Fails fell more than 30%: shown, never ranked | Set up safely on Hyperliquid → |
| 5 NYP- Not YOUR pool Public pool · Lighter · since Jan 13, 2026 | −66.4% | – | −43.9% | Fails fell more than 30%: shown, never ranked | Set up safely on Lighter → |
New listings (3)
All tools →Tools added to the directory this week, each from its official docs, with its facts and Safety score.
- Nov 17Meridian Engine: AI market analyses from Horizon AI (being incorporated in Italy) with a public record, sent by email, with no orders and no account access. The record is small: ten… Facts and Safety score →
- Nov 17Tritonix: an AI investment research agent from 3T Digital Limited (Hong Kong) with optional broker connectors. Its Copy Trade and Swing and Day Trade bots run in your own Alpaca,… Facts and Safety score →
- Nov 16Hashlock Markets: a non-custodial venue for cross-chain swaps between Bitcoin, Ethereum, Base, Robinhood Chain, TRON and Solana, where makers quote privately and each side locks its leg… Facts and Safety score →
Incidents this week (3)
Incident tracker →Security events for tools that touch money, each from the official disclosure.
- Nov 16Lighter, a perp DEX, said a security incident on Nov 16 came from a leaked API key for a specific trader's account: the affected accounts used their own signing code rather than Lighter's official SDK, and that code reused a… LighterIn the incident tracker Official source
- Nov 16Trezor said its shipping partner Skladon was breached on Nov 7 and an attacker exported customer shipping data for about 27,000 orders placed between Aug 9 and Nov 7, 2026: names, phone numbers, delivery and email addresses, and… In the incident tracker Official source
- Nov 16Report URI said malicious JavaScript served through Optimizely Web reached visitors of 178 organisations' websites on Nov 16, including ledger.com and OANDA's sites. The script swapped cryptocurrency wallet addresses on the page… In the incident tracker Official source
Affected? Follow what to do if a bot or wallet is drained, and put limits outside your bot with TAS Guard.
All weekly reports
- Week ending Nov 22, 2026 this pageWe checked what people using AI trading apps actually made. The typical user made nothing.Index 0.0% · 15 of 28 pass the gates · 3 incidents
- Week ending Nov 8, 2026Index 0.0% · 17 of 29 pass the gates · 3 incidents
- Week ending Nov 1, 2026Index +0.1% · 17 of 34 pass the gates · 7 incidents
Snapshot taken Nov 20, 2026 from data/perf/ (sources as of Nov 20, 2026). Vault and pool names are the operators' own; where one says it's AI-run, that's their claim, and the numbers are ours. Agents with a published account join when they pass the same gates: get yours verified.