In 2015 I formed a small group of engineers at Jane Street to rebuild the firm’s core trading system from the ground up, and we ended up cutting latency by two orders of magnitude. Some of the techniques we used, relevant for algorithmic trading systems and exchanges today:
Zero allocation: Whenever a program allocates memory for an object on the heap, the runtime pays a steep penalty in latency. The simplest solution is to avoid memory allocation entirely.
Jane Street famously uses OCaml, a strongly typed programming language that by default produces garbage collected by a dynamic collector. Most other firms use languages with manual memory management, but it was a strict part of Jane Street’s tech culture that all risk-sensitive code had to be written in OCaml. It took a collaborative effort across multiple groups within Jane Street’s technology org to create zero-allocation core libraries, combining the type safety of a functional programming language with the memory profile of a language like C.
We built the new main trading loop in this hybrid OCaml/C-style, producing zero new allocations in the critical path from tick to trade. In modern languages like Rust, it is substantially easier to achieve precise memory management while still benefiting from type safety and compile-time guarantees.
Kernel bypass: A primary goal of a low-latency trading system or exchange is to pull a network packet containing market data or order flow through the network card’s interface and into the program’s memory space as fast as possible. The standard Linux OS kernel uses slow abstractions to support a wide variety of network drivers, at the expense of the entire system’s end-to-end latency. When we started with an empty program that contained no business logic and only forwarded packets through when received, the end-to-end latency was already too slow.
To fix this issue, we employed a standard practice in the HFT industry in which we bypassed the OS’s kernel stack entirely by leveraging our network card vendors’ proprietary APIs to DMA packets straight from the NIC into memory. This technique brought our empty-packet-forwarding baseline into the latency regime we needed in order to build out the rest of the trading, risk, and protocol code.
Local IPC: Kernel bypass is necessary when reading routed packets off a network from a third party such as another exchange or client connection. When communicating between internal instead of external processes, the fastest transports avoid network stacks entirely.
Processes within the same box can transfer messages using shared memory or Unix domain sockets. This allowed us to continue with our familiar process boundaries for separable components without sacrificing significant performance. We had to write custom logic to emulate many of the features of network- and transport-layer protocols, with the result of creating a reusable, zero-overhead IPC mechanism.
Working on this problem was one of the most intellectually rewarding experiences of my early career. The above latency optimization techniques are fairly commonplace in the HFT trade but hard to learn outside the industry setting. Half of our team at Architect comes from Jane Street and other trading firms, and we value using our domain knowledge to build exchanges for the public rather than trading software that never leaves an HFT’s walls.
Show more
The protocols that route real revenue back to tokenholders are separating from the pack, and I think that's what the early stage of this altseason looks like.
$HYPE set the template by sending roughly 97% of fees into buybacks, well over $1B repurchased so far, and the token is up about 55% in the last 30 days. $LIT has done even better over the same window at +102%, after burning 6.3% of circulating supply in July and moving staking rewards off revenue onto an ecosystem reserve so that trading fees go entirely to reducing supply.
$PUMP is the reminder that buybacks alone don't carry a token. They put 100% of revenue into buybacks for nine months, burned $370M worth, and the token still spent most of the year below launch FDV before they cut the program to 50%. The mechanism works when there's growth underneath it.
$ENA is the clearest example of a team fixing alignment on purpose. The Foundation bought out the seed investors who had been selling, ended monthly VC unlocks with a single release in October, moved the protocol IP under tokenholder governance, and put a 5–20% revenue fee switch to a vote that turns on as USDe scales past $7.5B. Then they shipped Ethena Pay on Avalanche last week. $ENA is up 84% in a month.
ethereum:0xfe0c30065b384f05761f15d0cc899d4f9f9cc0eb is the largest DeFi-native neobank at roughly $100M a month in card spend, and it just passed a vote to make buybacks programmatic. A share of interchange, swap and staking revenue now gets TWAP'd into ethereum:0xfe0c30065b384f05761f15d0cc899d4f9f9cc0eb on chain every week, with 30% of each fee buying ETHFI for the Foundation treasury and another 30% buying ETHFI that goes back to the user as cashback. Two tokens bought for every one paid out, and rewards are funded by open-market purchases rather than new issuance.
ethereum:0x643c4e15d7d62ad0abec4a9bd4b001aa3ef52d66 is the one I think is most mispriced. Maple's AUM is $4.6B, up 81% year over year, buybacks went from discretionary to programmatic in August under MIP-021, and syrupUSDG is now live inside Robinhood Earn with access to 25M funded accounts. The token is still around 76% below its 2025 high.
What other alts are you watching that actually pay holders?
Show more
Statement on the $AXIS Sell-off of 28–29 October
Over the past two days, a series of addresses sold a large volume of $AXIS on-chain. We want to set out clearly what these tokens were and what we are doing about it.
1. No team or investor tokens were sold.
All Core Team and investor allocations, together with every other locked allocation, 835,000,000 $AXIS (83.5% of total supply), remain in the vesting contract, which is jointly administered with Binance: 0xAFE41cfd4C21f1D896678e32C5AD456ea41835e6. Anyone can verify the balance on BaseScan. Not a single token has left it, and none unlocks before the 12-month cliff.
2. Where the sold tokens came from.
The tokens sold came from the initial circulating supply set out in our published tokenomics, specifically part of the Ecosystem allocation. Many ecosystem recipients are not crypto-native, so their allocation was delivered through a single OTC counterparty, under a written agreement that includes lock-up terms.
3. What we believe happened.
The pattern of the sales, which went through exchange deposit routes and hundreds of fresh addresses within hours of delivery, strongly suggests that the OTC counterparty sold these tokens on the open market. If so, this is inconsistent with the agreement.
4. What we are doing.
- We have suspended all further dealings with this counterparty.
- We are reviewing the agreement with legal counsel and will pursue every remedy available under it.
- Going forward, any remaining Ecosystem tokens will be distributed only through on-chain vesting.
The team‘s tokens remain locked, and our commitment to building Axis is unchanged.
Show more
A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY.
Here's a full recap:
1. Nebius $NBIS got approval for its Vineland, New Jersey data center expansion, clearing a major hurdle tied to its $17.4B Microsoft $MSFT cloud deal. The Vineland Planning Board approved Phase 2 of the project, adding roughly 600,000 square feet and potentially increasing total campus capacity from about 300MW to around 350MW. DA Davidson had cut its price target from $250 to $175 last week over approval concerns, but raised it back to $250 after the decision, saying the approval removes a significant risk and allows Nebius to move forward with construction instead of becoming a “poster child” for data center delays.
2. Today’s selloff was driven largely by macro pressure. The VIX had been sitting near a two-year low, leaving stocks vulnerable to any pickup in volatility, while Treasury yields moved higher with the 10-year above 4.7% and the 30-year near 5.3%. Global bond pressure also intensified as Japan’s 30-year yield rose above 4% to a record high, while geopolitical uncertainty increased after Trump said the U.S. was not holding talks with Iran. The move was especially sharp in tech, with many semiconductor names down 8%-10%, the QQQs falling nearly 2%, and the S&P 500 down about 0.7%, as the market gave back part of the strong gains from the first half of August.
3. Anthropic is reportedly preparing to give CEO Dario Amodei and other co-founders enhanced voting rights ahead of a potential IPO as soon as late September, according to The Information. The structure is designed to protect leadership from outside shareholder pressure, especially as the founders hold relatively small economic stakes, with Amodei reportedly owning around 2%. Anthropic also plans to keep its Long-Term Benefit Trust, which holds special shares that allow it to elect a majority of the company’s board.
4. OpenAI said revenue reached $6.7B at the end of Q2, up 18% QoQ from $5.8B in Q1, while Anthropic reportedly doubled revenue to $11.6B over the same period. OpenAI’s operating loss widened from $9.3B in Q1 to $12.3B in Q2, outpacing revenue growth as compute costs, free-user subsidies, and pricing pressure weighed on margins. The slowdown matters because OpenAI has signed massive infrastructure deals tied to the expectation that it can eventually generate hundreds of billions in annual revenue, with companies like $NVDA, $ORCL, $CRWV, and $MU all exposed to the pace of AI demand. OpenAI is now leaning on products like Codex and a broader ChatGPT “super app,” while Greg Brockman has taken a more active role across product and business to help reaccelerate growth.
5. July economic data came in mixed. Import prices fell 0.4% MoM versus +0.1% expected, while export prices dropped 1.3% MoM versus +0.2% expected. Housing starts came in at 1.239M versus 1.350M expected, down 12.4% MoM, while building permits beat at 1.443M versus 1.375M expected. ADP also said private employers added an average of 9,500 jobs per week over the four weeks ending August 1, the first increase after 7 straight weeks of declines, but still well below the 30,750/week pace seen in early June.
6. The top 10 most active options today by contracts traded were $NVDA with 1.9M contracts, $TSLA with 1.5M contracts, $AAPL with 1.0M contracts, $INTC with 868K contracts, $MU with 847K contracts, $META with 823K contracts, $SPCX with 747K contracts, $AMZN with 700K contracts, $AMD with 385K contracts, and $NFLX with 359K contracts.
7. UBS remains bullish on Micron $MU with a Buy rating and $1,625 price target, arguing that cautious investor positioning creates an opportunity as earnings prove more resilient than feared. The firm says near-term concerns still center on supply agreement durability, gross margin downside, ongoing supply tightness, and future capacity additions, which could keep MU a battleground stock. However, UBS believes long-term investors may start to value memory more structurally as AI token generation economics become increasingly driven by memory rather than compute.
8. Onchain tokenized equity trading volume has hit a record $9B in 2026, up 207% QoQ and 800% year-to-date. The surge is being driven by demand for 24/7 global access to high-momentum stocks, especially memory and storage names, with Jupiter leading the Solana-based tokenized equities push and reporting 95% QoQ growth in routed volume. Jupiter also says 55% of tokenized equity volume happens during off-hours, while Nasdaq is moving in the same direction with plans to expand trading to 23 hours a day, five days a week.
9. Bank of America’s August fund manager survey shows investor bullishness approaching extreme levels, with a net 56% of managers overweight equities, the highest reading since November 2021. Cash allocations have dropped to just 3.5%, making this the third-most bullish survey since 2022. Despite strong confidence in stocks and AI spending, BofA says investors should consider rotating toward more defensive assets as market risks rise.
10. Einride is planning to deploy 500 Tesla $TSLA Semis for Amazon and other customers across major U.S. freight corridors beginning in September. The 24-month rollout will cover California, Texas, New Jersey, Illinois, and Georgia, and is expected to triple Einride’s deployed electric-truck fleet. The company operates electric and autonomous freight fleets through its Saga AI platform.
11. Google $GOOGL has reportedly told suppliers it plans to move production of all Pixel smartphones, watches, and earbuds out of China starting in 2027, with manufacturing shifting mainly to Vietnam and India. If completed, Google would become the second major global smartphone brand after Samsung to fully move smartphone production out of China. The company is also targeting 8%-10% Pixel shipment growth this year, up from roughly 12M units in 2025.
12. Morgan Stanley says active funds remain under-owned in mega-cap tech, with the gap widening to -129 basis points in Q2. Nvidia $NVDA is still the most under-owned large-cap tech stock, with its ownership gap widening 14 basis points to -2.53%, followed by Apple $AAPL at -2.33%, Microsoft at -1.54%, and Amazon $AMZN at -1.29%. Meanwhile, active managers are heavily overweight AI memory and storage names, with SanDisk $SNDK the most over-owned at +2.30%, while Lam Research $LRCX and Western Digital $WDC also rank among the most over-owned. Morgan Stanley says this shows a clear institutional bias toward AI “picks and shovels,” while under-owned mega-cap tech could benefit if active managers increase exposure.
WALL STREET IS THE GREATEST SHOW ON EARTH.
Show more