Battle Test: A Runaway Bot vs. the Risk Engine
On Saturday afternoon, a trading bot malfunctioned and briefly became most of the market in HYPE and ZEC on Lighter. HYPE printed $60.96 and $56.31 within half a minute of each other while the rest of the world traded it near $58. It was a real battle test of the risk engine, at size, on mainnet.
TL;DR: Everything worked as expected. Prices recovered within a minute, there were no liquidations across the market, and the losses were limited to the account that caused them. Full breakdown below. All numbers come directly from exchange data and were verified using two separate methods.
The Bot
The account was funded with $500k (deposited two hours before its first trade on July 15) and spent ten quiet days as a small taker bot: 12,028 fills, median fill ~$54, $3.5M total volume, always the taker (zero maker fills), zero fees paid.
Then, at 4:14 PM on Saturday, its sizing logic broke.
Five Minutes of HYPE
In five minutes the bot round-tripped $27.0M of HYPE, 74.2% of the entire market's volume. It flipped its position from +13.1k to −51.9k to +100.2k to −74.2k HYPE, buying its own slippage on every flip. The price briefly moved between $56.31 and $60.96 before returning to normal within a minute after the activity stopped. Over the same period, HYPE on Hyperliquid stayed between $57.83 and $58.13.
The bot then moved to ZEC and repeated the same pattern with larger size, trading $220.8M in taker volume, or 86.7% of the market, from 4:29 to 7:39 PM. ZEC briefly jumped from $478 to $521 on Lighter, while trading between $479.9 and $486.0 on Hyperliquid. After that, it traded smaller sizes across a few other markets until 12:30 AM Sunday, when it stopped trading.
Risk System Response
No liquidations: During both bursts, there were no liquidation or ADL fills for the bot or any other trader. On an exchange that marks positions using the last traded price, moves like these could have liquidated other users. Lighter instead marks positions using a manipulation resistant fair price based on the median of the order book impact price, the index price with a capped EMA premium, and CEX mark prices, so short lived price spikes like these do not affect anyone's margin.
No runaway losses: When the bot's account value fell below its initial margin requirement, the risk engine stopped letting it grow its position: 801 market orders were rejected (order status "canceled-margin-not-allowed") against 1,002 filled on July 25 alone. Margin checks run in the same verifiable circuits as order matching and liquidations. They are enforced on every transaction, with no discretion involved, so a malfunctioning bot can only burn its own collateral.
Where the Money Went
The bot lost ~$226k on HYPE and ~$218k on ZEC in the bursts alone. By the time it went flat, $482k of its $500k was gone, leaving $18k, all of it the operator's own funds. Lighter charges standard accounts zero fees, so the exchange took nothing.
On the other side, LLP (Lighter liquidity pool), earned about $143k across the two bursts, and 40 independent accounts, none affiliated with the bot or the protocol, each made more than $1k by providing resting limit orders that were filled in strict price time priority. The largest winner earned about $38k across HYPE and ZEC. Frontend data also showed traders using the new Chase Limit order type at roughly twice the normal rate during the event. One trader used it through both bursts and finished among the top five winners with about $13.7k .
Why This Matters
Writing a trading bot has never been easier. Neither has losing six figures in minutes to one line of bad sizing logic. The market structure did its job here: prices self-healed in a minute, nobody got liquidated, and the losses stayed on the account that caused them.
If you're building a bot, build on rails designed for it: Lighter's official Python and Go SDKs (with runnable examples), the API docs, and lighter-agent-kit for AI-agent trading. All of it is open source and runs on an exchange with an independent audits. The audit reports, including Nethermind's, are available in Lighter's docs. Start small, keep your order sizes in check, and remember: the margin engine is your last line of defense, not your first.
Be careful with your trading software. Stay safe out there.
Methodology: Fills, orders, and hourly account snapshots from Lighter's databases; PnL computed two independent ways (trade flows + window-end marks vs. protocol oracle marks + funding), agreeing within 0.05% (HYPE) and 0.7% (ZEC); prices cross-checked against Hyperliquid's public API; Chase Limit usage from frontend telemetry matched to exchange fills; funding trail verified on-chain (Ethereum).
75·AShort
m
meme7/24meme
reply @Thomus0x eh, Rosenblatt June 30 note identified AMD as the likely first $AAOI CPO customer. I wouldn't quite say they aren't remotely close.
$LITE / $COHR / $SIVE just look like your top 3 right now in terms of CPO laser visibility per the recent Morgan Stanley note.
75·ALong
m
meme7/23meme
$INTC and $AMD to sign CPU LTAs with Chinese customers for AI DCs (Reuters).
- Prices of some CPU products have risen more than 40% in China since the start of the year from sources.
- Month-on-month increases topping 10% for some products
CPUs were already a bottleneck, following CPU ratios due to AI inference...
But the broader trend of LTAs seems to be appearing from:
- Memory, with $MU, Samsung, $SNDK, and SK Hynix signing DRAM/NAND LTAs.
- Photonics, with $LITE, $COHR signing EML LTAs. And recent Trendforce reports that $AMD and hyperscalers are now pursuing CW LTAs.
And I'm sure there's many more from MLCCs to all the way to substrates.
+1 for the bottleneck investors... hard to be a "bubble that pops" if you have take or pay demand spanning multiple years.
75·ALong
m
meme7/22meme
quote: Tesla shareholder update underway right now! https://x.com/tesla/status/2080021924072366329 | Q2 Shareholder Update → ir.tesla.com
Highlights
– Cybercab began production at Gigafactory Texas
– Tesla Semi remains on track for volume production this year at our new factory in Nevada
– Making continued progress with battery pack capacity expansion (the main limiting factor to near-term vehicle production volume increase)
– Megafactory Texas is nearing completion (start of production planned for this year)
– More customers are now opting to subscribe to FSD at the time of vehicle purchase!
– Robotaxi rollout continued in the US. Now live in 7 major metros
– Construction of Optimus at Fremont Factory began after decommissioning the Model S & X lines. Planned production later this year
From here, there remains much hard work as we aim to revolutionize transportation, energy and productivity through our leading real-world AI. Scaling will be non-linear and we are focused on long-term value creation.
We’ve never been more optimistic about the future.
Automotive
– Record deliveries in several markets: South Korea, Australia, Colombia, Japan, Taiwan, Thailand, Portugal, the Philippines, Chile, Slovenia & Lithuania
– We launched the Model YL in the US in July and have seen a positive response from customers
Energy generation and storage
– Record energy storage deployments in EMEA, supported by record deployments from Megafactory Shanghai, which continues to ramp production
– On track to begin production of Megapack 3 & Megablock this year at our new Megafactory Texas
– Powerwall 3P (three-phase) is now available in Germany and is designed to meet the power needs of German homes with a single unit
Robotics
– Installing the first-generation lines for Optimus at Fremont Factory, where we expect to start production soon
– The initial Optimus builds will be used in our Optimus Academy for training data collection and further functionality development
– Additionally, we continued site development at Gigafactory Texas with building construction now in full swing
AI Training Compute
– More than doubled our onsite compute in Texas (in terms of MW of compute) during the first half of 2026
– Cortex 2 supports the development of both vehicle and humanoid robot autonomy software & will ramp further over the rest of the year
Battery
– Ramping new battery & material factories, including vehicle pack capacity in Berlin, cathode material production and lithium refining in Texas and LFP cells in Nevada for our energy storage products
– Increasing production of 4680 cells to support ramping Cybercab & Tesla Semi plus increased production of Model Y
Other Supporting Infrastructure
– Added over 2,400 net new Supercharging stalls, growing the network by 17% year-over-year
AI Software
– Started rolling out FSD v14 lite to early-access customers in the US & South Korea with AI3 hardware
This software build distills the driving behavior from AI4’s v14 series into both the camera & compute configuration of AI3, bringing destination options & speed profiles. It also addresses challenging driving scenarios with improved proactive & reactive responsiveness
AI Inference Compute
– Making progress on construction & equipment procurement for our semiconductor fab in Austin
Automotive and Other Software
– Rolled out Summer Release:
Self-Driving stats are now available in the mobile app
Grok can make phone calls, search & play music, and adjust climate controls, among other things
Automatic Navigation expands beyond home & work to support any destination based on personal habits and schedule
Robotaxi
– Started production of Cybercab, our purpose-built autonomous EV designed to be the workhorse of our Robotaxi fleet
– Began offering employee rides in Cybercabs on our GFTX campus in July
– Preparing for expansion of our Robotaxi service to additional US metros: testing, permitting & first responder training
– Expanded unsupervised rides to the entire Austin metro area & launched unsupervised rides in Miami, Orlando & Tampa in July
FSD Supervised
– Record net new subscriptions in Q2
– Record FSD attach rates in North America, with over half of new deliveries including FSD subscriptions
– Received additional approvals in Lithuania, Estonia, Denmark & Belgium, with customers in these countries driving over 50 million kilometers (31 million miles) on FSD as of July
80·ALong
m
meme7/22meme
$GOOGL How much are they earning from token sales?
An estimate:
Per the call, GOOGL AI model APIs are processing ~22B tokens per minute (+38% q/q; ~270% annualized), with 500 cloud customers >1 trillion in LTM and 2k enterprises >100B in LTM. Existing customers are exceeding their commitments by >50%.
Rough sizing attempt with Claude on this:
Assuming 22B tokens per minute (implies 11.6 quadrillion per year run-rate), and pricing between $0.10 and $2.00 per million, would imply revenue between $1.2B - $23B (likely on lower half range, considering this skews to cheap flashlite/free tier models AND includes Google's own internal consumption, which is included in token count but NOT in Cloud revs).
Also probably makes more sense to only assume ~70% of these tokens have any monetization.
This analysis shows how powerful the opportunity is as 1) token sales continue to grow high 30s% and 2) they are able to capture more pricing (more off free/lite tiers).
AI tokens sales are likely only 5-15% (low-end) of annualized Cloud revs.
h/t Griffin MacMaster
75·ALong
m
meme7/22meme
quote: $GOOGL now has 950M monthly Gemini users and processes 22B API tokens/min.
Compared to 750M back in February.
Absurd growth adding 200 million more users in a few months. Kinda explains why they ran out of compute, then cut allocations to $META and others.
Very unlikely AI capex will slow down given. | $GOOGL reported earnings today:
EPS: $9.11 actual vs $2.91 expected - retail reactions to EPS blowout is kinda just accounting noise, since $SPCX, Anthropic, and others were likely large contributors.
Revenue: $119.7B actual vs. $116.98B expected
Google Cloud Revenue: $24.77B +82% Y/Y, vs. 63-65% expected ($22.2-$22.6B) - this is a very large beat and most material part so far.
Cloud Operating Income: ~$8.81B, implying 35.6% margin vs. 32.9% Q1.
Capex: $44.92B vs ~$44.15B (basically in line).
So initial read through is Google Cloud accelerating growth with expanding margins is genuinely bullish for AI demand.
Their former $180-$190B capex guidance is already extremely large, and as long as we get around these numbers in the earnings call + forward projections...
Should be good to go for $LITE and the other optics/networking trade. (eg. last earnings, they said DC and networking would be ~40% of capex spend).
Hyperscaler earnings transcripts are probably the most important thing to pay attention to with the AI capex trade.
And that should be in 3 minutes.
80·ALong
m
meme7/22meme
$GOOGL reported earnings today:
EPS: $9.11 actual vs $2.91 expected - retail reactions to EPS blowout is kinda just accounting noise, since $SPCX, Anthropic, and others were likely large contributors.
Revenue: $119.7B actual vs. $116.98B expected
Google Cloud Revenue: $24.77B +82% Y/Y, vs. 63-65% expected ($22.2-$22.6B) - this is a very large beat and most material part so far.
Cloud Operating Income: ~$8.81B, implying 35.6% margin vs. 32.9% Q1.
Capex: $44.92B vs ~$44.15B (basically in line).
So initial read through is Google Cloud accelerating growth with expanding margins is genuinely bullish for AI demand.
Their former $180-$190B capex guidance is already extremely large, and as long as we get around these numbers in the earnings call + forward projections...
Should be good to go for $LITE and the other optics/networking trade. (eg. last earnings, they said DC and networking would be ~40% of capex spend).
Hyperscaler earnings transcripts are probably the most important thing to pay attention to with the AI capex trade.
And that should be in 3 minutes.
75·ALong
m
meme7/21meme
quote: Swarms x @GeminiApp
Gemini 3.6 Flash and the latest Gemini models are now available on Swarms Cloud, making it easy to build production-ready single-agent and multi-agent workflows.
This guide covers API setup, single-agent execution, and multi-agent orchestration with complete Python and cURL examples so you can start building with Gemini in minutes.
Learn more below ⬇️ | We’re rolling out three new models to make AI agents faster, smarter, and cheaper at scale:
🔵 Gemini 3.6 Flash: It uses fewer tokens than 3.5 Flash to deliver higher quality work at the exact same cost.
🔵 Gemini 3.5 Flash-Lite: A fast, cost-effective option for everyday tasks like processing documents and agentic search.
🔵 Gemini 3.5 Flash Cyber: A cybersecurity model built to find and patch critical software vulnerabilities.
75·ALong
m
meme7/21meme
quote: Swarms x @GeminiApp
Gemini 3.6 Flash and the latest Gemini models are now available on Swarms Cloud, making it easy to build production-ready single-agent and multi-agent workflows.
This guide covers API setup, single-agent execution, and multi-agent orchestration with complete Python and cURL examples so you can start building with Gemini in minutes.
Learn more below ⬇️ | We’re rolling out three new models to make AI agents faster, smarter, and cheaper at scale:
🔵 Gemini 3.6 Flash: It uses fewer tokens than 3.5 Flash to deliver higher quality work at the exact same cost.
🔵 Gemini 3.5 Flash-Lite: A fast, cost-effective option for everyday tasks like processing documents and agentic search.
🔵 Gemini 3.5 Flash Cyber: A cybersecurity model built to find and patch critical software vulnerabilities.