Industries / Finance and business services / Finance and banking
Trading and markets
Markets are the places, now mostly computer systems, where shares, bonds and contracts based on them are bought and sold, and where each trade is then settled, meaning the cash and the securities change hands.
7
Milestones achieved, from 1602 to 2023.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1602
Shares anyone could trade
The Dutch East India Company (VOC), founded in 1602, sold shares that could easily pass from one owner to another, and buyers began trading them almost at once. The publisher's summary of historian Lodewijk Petram's book on that market says traders there went on to use forwards, futures and options, contracts to buy or sell later at a price agreed now. Guinness World Records lists the Amsterdam Stock Exchange, founded in 1602 for dealings in the VOC's printed shares, as the oldest stock exchange.
Credit
the Dutch East India Company; Lodewijk Petram for the history
1973
An exchange for options
An option is a contract that gives the right, but not the duty, to buy or sell a stock at a set price before a set date. On 26 April 1973 the Chicago Board Options Exchange (Cboe) opened, offering standardized option contracts, with the same terms for every buyer, on a national securities exchange, and with next-day settlement. It began with call options, the right to buy, on 16 stocks, while many on Wall Street doubted it would survive, by the account of Financial History magazine.
Credit
the Chicago Board Options Exchange
1977
Computers take the orders
The Toronto Stock Exchange launched its Computer Assisted Trading System (CATS) in 1977, which TMX Group, the exchange's owner, calls the world's first. Between 1995 and 1997 Toronto became the largest exchange in North America to move to a floorless, electronic market, by TMX's account. The New York Stock Exchange ran without its trading floor for the first time on 23 March 2020, during COVID-19, and partly reopened it on 26 May 2020.
Credit
the Toronto Stock Exchange
1987
Black Monday
On 19 October 1987 the Dow Jones Industrial Average, a widely followed US stock index, fell 508 points, or 22.6%, its largest one-day percentage fall, by the Federal Reserve's history site. Portfolio insurance, a hedging method that sold futures as prices fell, added to the selling. The Federal Reserve said the next day that it was ready to serve "as a source of liquidity", meaning cash for the financial system, and the New York Stock Exchange later adopted circuit breakers, which halt trading for a time after a large fall in the S&P 500 index.
LessonSelling rules that react to falling prices can feed the fall they were meant to guard against.
Credit
the Federal Reserve; the New York Stock Exchange, for circuit breakers
2010
The flash crash
On the afternoon of 6 May 2010 major US stock indexes, already down more than 4%, fell a further 5–6% in minutes and recovered almost as fast. More than 20,000 trades in over 300 securities were executed at prices more than 60% away from their values moments before, some at a penny or less and some at $100,000. The report of the staffs of the SEC and the Commodity Futures Trading Commission (CFTC) describes a fund manager selling 75,000 E-Mini contracts, futures on the S&P 500 index worth about $4.1 billion, through an algorithm set to trade 9% of volume "without regard to price or time", which finished the sale in 20 minutes. Regulators then added circuit breakers for single stocks, which pause trading in a stock for five minutes after a 10% move within five minutes.
LessonIn the report's words, "high trading volume is not necessarily a reliable indicator of market liquidity".
Credit
SEC and CFTC staff, for the report
2012
Knight Capital's 45 minutes
On 1 August 2012 the US broker Knight Capital deployed new code to its automated order router incorrectly, which set off an old function it had left in the router but did not mean to use. In the first 45 minutes after the market opened, the router sent more than 4 million orders while trying to fill 212 customer orders, traded more than 397 million shares and lost Knight more than $460 million, by the SEC's account. The SEC fined Knight $12 million for breaking its market access rule, Rule 15c3-5, which governs the risk checks a broker must apply to orders it sends to an exchange.
LessonUnused code left in a live system turned one faulty software release into a loss of more than $460 million.
Credit
the SEC, for the enforcement case
2023
Settling in one day
In January 2021 GameStop's shares closed at $347.51 on 27 January, more than 1,600% above their 11 January close, and in late January brokers including Robinhood and Interactive Brokers let customers only sell, not buy, some such stocks. The SEC staff report on those weeks said the episode "raises questions about the possible effects of acute margin calls", the clearinghouse's demands for deposits, on thinly capitalized brokers, and that one way to reduce the risk "is to shorten the settlement cycle". India finished moving all stocks to settlement one business day after the trade (T+1) on 27 January 2023. The US followed on 28 May 2024, under an SEC rule adopted on 15 February 2023 that answered "the meme stock events of 2021".
LessonThe longer trades wait to settle, the more a clearinghouse asks brokers to deposit against them, and on volatile days that demand can force brokers to limit trading.
Credit
India's market regulator, SEBI, and its stock exchanges; the SEC
Watch for
EU moves to T+1
Regulation (EU) 2025/2075, adopted on 8 October 2025, amends the EU's rules for central securities depositories, the bodies that hold securities and settle trades, so that EU trades in transferable securities settle one business day after the trade from 11 October 2027.
Achieved whenEU trades settle at T+1 from that date and ESMA, the EU's markets authority, or a national regulator publishes settlement-failure figures from after the switch.
Claimed by
the European Union, October 2025
Tokenized shares at DTC
A tokenized share is an ordinary share whose ownership is also recorded on a blockchain, a shared digital ledger. In December 2025 SEC staff gave the Depository Trust Company (DTC), which holds and settles most US securities, a no-action letter, a statement that staff would not recommend enforcement, for a service that makes digital versions of Russell 1000 stocks, ETFs on major indexes and US Treasury securities available on approved blockchains for an initial three years. DTC expected the service to go live in the second half of 2026. On 18 March 2026 the SEC approved a Nasdaq rule that lets participants in DTC's pilot have stock trades settled in tokenized form, with tokenized shares on the same order book as ordinary ones.
Achieved whenTokenized trades settle in routine production and an outside party, such as the SEC or an auditor, reports volumes and settlement failures.
Claimed by
DTCC, December 2025, and Nasdaq, March 2026
Open
01
Read corporate action notices
A public benchmark for reading corporate action notices, the announcements of dividends, stock splits, mergers and tender offers that change what a shareholder owns. Collect notices from public filings, label by hand the event type, ex-date (the first day a buyer no longer gets the dividend), record date, payment date, ratio or amount, and any choices offered to holders with their deadlines, and score extraction methods field by field. Only public documents are used.
The briefCredit · unclaimed02
Test cases for abuse alerts
An open library of synthetic trading scenarios for testing the market-abuse alerts that brokers run, such as insider dealing ahead of a price-sensitive announcement, and pump and dump, where a price is pushed up so the pusher can sell. Each case is planted in simulated normal trading for stocks with low, medium and high price swings, and labeled. Run a set of open baseline alerts over it at several thresholds and publish what each catches and misses. All data is simulated, and no real accounts or trades are used.
The briefCredit · unclaimed
Open 01 · the brief
Read corporate action notices
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public benchmark for reading corporate action notices, the announcements of dividends, stock splits, mergers and tender offers that change what a shareholder owns. Collect notices from public filings, label by hand the event type, ex-date (the first day a buyer no longer gets the dividend), record date, payment date, ratio or amount, and any choices offered to holders with their deadlines, and score extraction methods field by field. Only public documents are used.
Why it is still open
Straight-through processing, handling with no person stepping in, has fallen to between 50% and 70% in asset servicing, the work that includes these events, by an industry survey DTCC co-sponsored in 2024, which says the gap means "additional manual intervention". A public dataset built from SEC filings notes that there is "no free source" for ex-dates, record dates or pay dates. The two public extraction projects found have no measured accuracy: review-radar has a hand-labeled set of 66 SEC 8-K filings and leaves its model column "unmeasured", and lexis-agent tested 8 Brazilian notices with no answer key.
Built with what exists
SEC EDGAR 8-K filings and their attached press releases; DTCC's published ISO 20022 corporate action message specifications as the target format; a language model to read notices, such as Claude Fable 5.2 (see the Note "Three working devices") or the open-weight Mistral Large 4 (see the Note "A trillion open parameters"); Gemini Decision 1 to sort notices by event type (see the Note "A thousand questions, one pass"); a spreadsheet for two-person labels.
First step
Pull 50 8-K filings that announce dividends or splits from EDGAR, label them by hand, and compare the labels with review-radar's 66-filing set.
Done looks like
At least 500 notices from at least two countries, each labeled by two people with disagreements settled and recorded, released under an open license, and a published table of field-by-field accuracy for at least two methods, one of them a simple rules-based baseline, that someone else can rerun from the released code.
Open 02 · the brief
Test cases for abuse alerts
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open library of synthetic trading scenarios for testing the market-abuse alerts that brokers run, such as insider dealing ahead of a price-sensitive announcement, and pump and dump, where a price is pushed up so the pusher can sell. Each case is planted in simulated normal trading for stocks with low, medium and high price swings, and labeled. Run a set of open baseline alerts over it at several thresholds and publish what each catches and misses. All data is simulated, and no real accounts or trades are used.
Why it is still open
In October 2026 the FCA, the UK regulator, reported that at seven retail brokers alert testing "was often limited", and that some used "a flat price move threshold" for insider-dealing alerts without checking it against stocks of different volatility. It named testing with synthetic trade data as good practice. The one recent synthetic market-abuse task found, a 1,000-account pump-and-dump set in an October 2026 paper, does not say its data is released.
Built with what exists
StockSim, an open-source market simulator that models individual orders, released in July 2025 (paper); Python; an AI coding agent to write scenario generators, with a person checking each scenario's label.
First step
Write ten insider-dealing scenarios around a made-up announcement, generate background trading in StockSim, and compare a flat 10% price-move alert with one scaled to each stock's usual volatility.
Done looks like
At least 200 labeled scenarios across at least three abuse types and three volatility bands, released openly, and a published table giving, for each baseline alert at three threshold settings, the share of abuse cases caught and the false alerts per 10,000 normal simulated accounts.