Another went to tape backups. A third went to a paper printout, the ultimate but most inconvenient record-keeping modality. The nature of the interfaces demanded that each output come from a different internal board inside the computers, but they were all the same output, and as a result nobody bothered with the permanent records. After all, there were a total of six machines divided between two separate locations. This system was as secure as people could make it.

Things could have been done differently. Each sale/purchase order could have been sent out immediately, but that was untidy-the sheer administrative volume would have taxed the abilities of the entire industry. Instead, the purpose of DTC was to bring order out of chaos. At the end of each day, the transactions were organized by trading house, by stock issue, and by client, in a hierarchical way, so that each house would write a limited number of checks—funds transfers were mostly done electronically, but the principle held. This way the houses would both save on administrative expense and generate numerous means by which every player in the game could track and measure its own activity for the purposes of internal audit and further mathematical modeling of the market as a whole. Though seemingly an operation of incomprehensible complexity, the use of computers made it as routine and far more efficient than written entries in a passbook savings account.

"Wow, somebody's dumping on Citibank," the sys-con said.

The floor of the New York Stock Exchange was divided into three parts, the largest of which had once been a garage. Construction was under way on a fourth trading room, and local doomsayers were already noting that every time the Exchange had increased its space, something bad had happened. Some of the most rational and hard-nosed business types in all the world, this community of professionals had its own institutional superstitions. The floor was actually a collection of individual firms, each of which had a specialty area and responsibility for a discrete number of issues grouped by type. One firm might have eight to fifteen pharmaceutical issues, for example. Another managed a similar number of bank stocks. The real function of the NYSE was to provide both liquidity and a benchmark. People could buy and sell stocks anywhere from a lawyer's office to a country-club dining room. Most of the trading in major stocks happened in New York because …it happened in New York, and that was that. The New York Stock Exchange was the oldest. There were also the American Stock Exchange, Amex, and the newer National Association of Securities Dealers Automatic Quotation, whose awkward name was compensated for by a snappy acronym, NASDAQ. The NYSE was the most traditional in organization, and some would say that it had been dragged kicking and screaming into the world of automation. Somewhat haughty and stodgy—they regarded the other markets as the minor leagues and themselves as the majors—it was staffed by professionals who stood for most of the day at their kiosks, watching various displays, buying and selling and, like the trading houses, living off the "middle" or "spread" positions which they anticipated. If the stock market and its investors were the herd, they were the cowboys, and their job was to keep track of things, to set the benchmark prices to which everyone referred, to keep the herd organized and contained, in return for which the best of them made a very good living that compensated for a physical working environment which at best was chaotic and unpleasant, and at its worst really was remarkably close to standing in the way of a stampede. The first rumblings of that stampede had already started. The sell-off of Treasury notes was duly reported on the floor, and the people there traded nervous looks and headshakes at the unreasonable development. Then they learned that the Fed had responded sharply. The strong statement from the chairman didn't—couldn't—disguise his unease, and would not have mattered in any case. Few people listened to the statement beyond the announcement of a change in the Discount Rate. That was the news. The rest of it was spin control, and investors discounted all of that, preferring to rely on their own analysis.

The sell orders started coming. The floor trader who specialized in bank stocks was stunned by the phone call from Columbus, but that didn't matter.

He announced that he had "five hundred Citi at three," meaning five hundred thousand shares of the stock of First National City Bank of New York at eighty-three dollars, two full points under the posted price, clearly a move to get out in a hurry. It was a good, attractive price, but the market hesitated briefly before snapping them up, and then at "two and a half."

Computers also kept track of trading because the traders didn't entirely trust themselves to stay on top of everything. A person could be on the phone and miss something, after all, and therefore, to a remarkable degree, major institutions were actually managed by computers, or more properly the software that resided on them, which was in turn written by people who established discrete sets of monitoring criteria. The computers didn't understand the market any more than those who programmed them, of course, but they did have instructions: If "A" happens, then do "B." The new generation of programs, generically called "expert systems" (a more attractive term than "artificial intelligence") for their high degree of sophistication, were updated on a daily basis with the status of benchmark issues from which they electronically extrapolated the health of whole segments of the market. Quarterly reports, industry trends, changes in management, were all given numerical values and incorporated in the dynamic databases that the expert systems examined and acted upon, entirely without the judgmental input of human operators.

In this case the large and instant drop in the value of Citibank stock announced to the computers that they should initiate sell orders on other bank stocks. Chemical Bank, which had had a rough time of late, the computers remembered, had also dropped a few points in the last week, and at the three institutions that used the same program, sell orders were issued electronically, dropping that issue an instant point and a half. That move on Chemical Bank stock, linked with the fall of Citibank, attracted the immediate attention of other expert systems with the same operational protocols but different benchmark banks, a fact that guaranteed a rippling effect across the entire industry spectrum. Manufacturers Hanover was the next major bank stock to head down, and now the programs were starting to search their internal protocols for what a fall in bank-stock values indicated as the next defensive move in other key industries.

With the money realized by the Treasury sales, Columbus started buying gold, both in the form of stocks and in gold futures, starting a trend from currency and into precious metals. The sudden jump there went out on the wires as well, and was noted by traders, both human and electronic. In all cases the analysis was pretty much the same: a sell-off of government bonds, plus a sudden jump in the Discount Rate, plus a run on the dollar, plus a building crash in bank stocks, plus a jump in precious metals, all combined to announce a dangerous inflationary predictor. Inflation was always bad for the equities market. You didn't need to have artificial intelligence to grasp that. Neither computer programs nor human traders were panicking yet, but everyone was leaning forward and watching the wires for developing trends, and everyone wanted to be ahead of the trend, the better to protect their own and their clients' investments.

By this time the bond market was seriously rattled. Half a billion dollars, dumped at the right time, had shaken loose another ten. The Eurodollar managers who had been called back to their offices were not really in a fit state to make rational decisions. The days and weeks had been long of late with the international trade situation, and arriving singly back at their offices, each asked the others what the devil was going on, only to learn that a lot of U.S. Treasuries had been sold very short, and that the trend was continuing, now augmented by a large and very astute American institution. But why? they all wanted to know. That set them to looking at additional data on the wires, trying to catch up with the information streaming from America. Eyes squinted, heads shook, and these traders, lacking the time to review everything, turned to their own expert systems to make the analyses, because the reasons for the swift movements were simply not obvious enough to be real. But it didn't really matter why, did it? It had to be real. The Fed had just gone up a full point on the discount rate, and that hadn't happened by accident. For the moment, they decided, in the absence of guidance from their governments and central banks, they would defer buying U.S. Treasuries. They also began immediate examinations of their equity holdings, because stocks looked as though they were going to drop, and drop rapidly.