How a 6-Week OS Became the Foundation of Microsoft's Empire

The article references a previous piece about Gary Kildall, who accused DOS of "brazen theft" by printing DOS and CP/M system calls side by side. Many portray Tim Paterson as an opportunist who sold crude copied code to Gates for a few thousand dollars, who then licensed it to IBM to become the world's richest man. But that is not what happened in Seattle in 1980.

Chips Piling Up in the Warehouse

Paterson worked at Seattle Computer Products (SCP), boss Rod Brock. In 1979 Intel released the 16-bit 8086; fresh graduate Paterson built an S-100 bus 8086 board in under two months, a year before the IBM PC project. But the board had no OS. The microcomputer ecosystem was tied to 8-bit CP/M. Brock asked Digital Research in California when 16-bit CP/M-86 would ship; delayed from late 1979 to April 1980, still no delivery. The small company could not wait; cash was tied up in chips, and it would go under. In mid-April 1980, Paterson told his boss: stop waiting for California, I will write one myself.

Six Weeks Working from Public Manuals

Paterson's goal was practical, not a revolutionary architecture: just read and write floppies and run old software. If old 8080 business programs could not run, no one would buy a bare board. He had no CP/M source code, no decompiler, only the publicly published "CP/M Programmer's Reference Manual". The manual listed system call conventions: function 1 read console, 2 write character, 15 open file. Paterson used 8086 assembly to implement the low level while keeping the same input register conventions, so old software could run with just a translation script. About six weeks later the kernel ran. He named it QDOS 0.10 (Quick and Dirty Operating System). By August the disk driver was done, the boss renamed it 86-DOS and shipped it. Inventory sold out, and the small Seattle shop survived.

Two Checks from Across the Lake

Across Lake Washington in Bellevue, Bill Gates was also cornered. IBM handed Microsoft the entire PC software package. Gates hoped IBM would buy Kildall's CP/M, but after negotiations collapsed, IBM issued an ultimatum: deliver an OS or the contract would be void. Microsoft had rented SCP machines in late 1979 to debug 8086 BASIC. Paul Allen called the SCP boss directly. Microsoft's acquisition strategy was precise: in December 1980, it paid $25,000 for a non-exclusive license to 86-DOS, on the pretext of a confidential client; in May 1981, it hired Paterson full-time for porting; on July 27, 1981, two weeks before the IBM launch, Microsoft paid another $50,000 to buy all rights to 86-DOS. The two checks totaled $75,000. Brock thought he had struck a bargain; he did not even know the real buyer was IBM.

Writing Your Own Kernel by Reading Someone Else's Manual

For decades people claimed QDOS was just a CP/M reskin. But examining the code shows the kernels are completely different. CP/M was designed for 8-inch floppies and relied on the fragile FCB (File Control Block) mechanism; a write interruption often corrupted the entire floppy directory. Paterson hated that logic. He drew inspiration from Microsoft BASIC and wrote a centralized linked-list structure from scratch: the FAT (File Allocation Table) that dominated PCs for decades. Under FAT, the disk is divided into clusters; floppy read/write efficiency improved dramatically, and sudden power loss would not destroy the whole disk. Even daily commands changed: CP/M used the counter-intuitive PIP command; Paterson replaced it with the intuitive DIR, COPY, TYPE. The 20-plus identical system call numbers were his adapter for cross-platform compatibility; he wrote a clean compatibility layer, not a direct code copy.

From $40 to a Defamation Lawsuit

In August 1981, the IBM PC launched, with optional PC-DOS priced at $40. Six months later, Kildall's late CP/M-86 carried an unbelievable price of $240. Buyers only cared which system could run spreadsheets more cheaply; the shipment ratio quickly widened to 100:1. Kildall bought an IBM PC and disassembled the kernel. Finding that the first 26 system calls used the same input registers as CP/M, he trembled with rage. In his memoir he denounced: "I asked him why he did it. Paterson replied he just copied CP/M's interface... This is brazen theft." Kildall wanted to sue, but 1981 law had not clarified whether APIs could be copyrighted. IBM legal mediated, agreeing to sell CP/M-86 alongside. Kildall thought side-by-side users would recognize quality; he lost completely. The dispute extended past Kildall's death. In 2004, scholar Harold Evans in a bestseller accused Paterson of stealing CP/M. Paterson sued for defamation. In court, Paterson presented the derivation of the FAT file system, proving he had never seen CP/M source code. Three years later the court dismissed the suit—not because it found he copied, but because the author relied on a public memoir, not "actual malice." The court gave no technical verdict; the originality debate remains unresolved.

Who Got the Code Running Before the Deadline

In 1986, after Brock learned the truth, he sued Microsoft for fraud; Microsoft settled with a $1 million check on the eve of its IPO. That money gave the Seattle shop a windfall, but the same year Microsoft went public and Gates entered the ten-billion-dollar club. Paterson later worked on early Visual Basic. Facing decades of repeated questioning, his answer was always calm: "Did I plagiarize CP/M? No. I read their public manual line by line and wrote my own code." Setting aside moral judgment, Gary Kildall deserves sympathy: he opened the door to microcomputer operating systems but missed the delivery window due to arrogance and delay. But painting Paterson as a thief is equally unfair. He was not a business strategist, just a young engineer backed into a corner. Facing a warehouse full of unsold boards, a boss who could not pay salaries, and a six-week deadline, he picked up a public manual and used a rewritten FAT to carve a survival path for his company. The commercial world is sometimes that ruthless. Ideas and architectures are great, but the one who takes everything is always the person who gets the code running before the deadline.