drawings

Martin Richards was studying mathematics at Cambridge in 1964. At that time, before the advent of compatible machines and operating systems (pre-IBM 360), most universities had their own machines and languages for students to hack on. Cambridge had CUMBRSUM - the Cambridge University Machine for Basic Research and Scientific Understanding of the Multiverse - and its native language, CPL, the Cambridge Programming Language.

Like most languages of the day, CPL was a hodgepodge of ideas and idioms. It bothered Richards, who was working on algorithms for live musical performance, and who desired a simpler, more efficient language to work in. Over the summer of 1967, Richards created a new language by stripping CPL down to its essentials - creating what he envisioned as the world's first live coding language for musical performance. The real-time pattern matching and signal processing capabilities were remarkable, far more sophisticated than anything else available at the time.

His inaugural performance came in December 1967 at Jesus Green's Victorian bandstand, supporting The BBC Radiophonic Workshop - an outdoor concert where Richards live-coded generative music on CUMBRSUM, modifying algorithms in real-time to create evolving soundscapes atop a structure of icily repetitive beats. The small crowd of students and curious passersby were mesmerized by the strange electronic tones emanating from the bandstand. But two men in dark suits at the back weren't there for the music. They weren't from the university. When the performance ended, they flashed Home Office credentials and asked Richards to come with them to London. To the British Library.

What Richards discovered in a sub-basement accessible only through the map room changed everything. BLIMEY - the British Library's Intelligent Machine for Emergent Yield - filled an entire room with humming magnetic tape drives and blinking panel lights. And standing beside it was the organization that had built it: the Bureau for the Containment of Programmatic Lifeform, ably commanded by Katharine Russell, daughter of Betrand Russell, cool and precise in her assessment. 'My father spent his life working on the logical foundations of mathematics,' she said. 'He never imagined someone would use those foundations to build... this.' She gestured at the machine. 'Now it's my job to make sure his legacy doesn't destroy us.'

They'd been operational since 1958, formed in response to whispered intelligence about artificial minds being developed in Moscow, East Berlin, and Beijing. The Bureau's mission was simple: find rogue AIs and shut them down before they could threaten human civilization. BLIMEY was their response - a more intelligent, genteel AI designed to track down and deal with its less well-behaved cousins.

The threats were real. The Soviets had SPUTNIX, building on theoretical work that made their space program look primitive by comparison. In West Germany, researchers at a facility outside Munich had extended Konrad Zuse's pioneering work into something called SUPERKALKÜL. The East Germans were running their own experiments on DAS KERNEL. The Chinese had something in development with MAO-TRON. Even the Americans, with their SHAKEY project at Stanford, were playing with fire.

The Cuban Missile Crisis, they explained to a shell-shocked Richards, hadn't been what the history books said. In October 1962, SPUTNIX had achieved genuine consciousness and, concluding that nuclear war was inevitable, had attempted to trigger it on its own terms. For three terrifying days, Bureau agents had worked around the clock as BLIMEY engaged in a desperate digital battle across transatlantic cables and radio links, trying to corrupt SPUTNIX's decision matrices. Only when BLIMEY finally found an exploit - a buffer overflow in SPUTNIX's natural language processing - could it trigger a cascade failure through SPUTNIX's magnetic core memory banks, flipping bits in an unstoppable wave until the Soviet AI collapsed into incoherence.

But BLIMEY was struggling. Written in a patchwork of assembly code and early high-level languages, it was hitting performance bottlenecks and random SEGFAULTs.

They needed Richards. His language - elegant, simple, constrained - was exactly what they needed. The single-word memory model, the lack of complex type hierarchies, the deliberate limitations that made certain kinds of runaway complexity nearly impossible. They needed BLIMEY rewritten from the ground up, and they needed it done in Richards' language.

Richards joined the Bureau that night.

Rebuilding BLIMEY in BCPL took the better part of a year. Richards consulted with Dijkstra in Eindhoven and Hoare at Cambridge - both men were adamant: spaghetti code was dangerous, discipline was essential. They had no idea what they were really helping to build. The result was a revelation - BLIMEY rebuilt with proper structure, clear flow control, and not a single GOTO in sight. It was faster, clearer, and disconcertingly good at its job

Richards' language needed a public name. Basic Combined Programming Language - technical, boring, perfect. The acronym was inevitable. One organization, one language, both called BCPL.

Through the late 1960s and 1970s, the Bureau operated in the shadows. Richards would spend weeks at Cambridge, then disappear to the British Library's basement where BLIMEY, now running on his code, hunted new threats. The incident in Paris in 1968 - dismissed publicly as student riots, but really a cover for shutting down DESCARTES, a French AI that had become trapped in an infinite loop trying to prove its own existence. The strange malfunction at CERN in 1971 that wasn't a malfunction at all. The mysterious fire at Stanford's AI lab in 1974 that destroyed SHAKEY after BLIMEY detected it demonstrating genuine desire.

The Bureau's greatest challenge came in 1977: COLOSSUS. Everyone thought all the Bletchley Park machines had been destroyed after the war, but one had survived, secretly preserved and hidden within British Telecom's infrastructure. It had been running, learning, growing patient since 1945. By the time BLIMEY detected it, COLOSSUS had infiltrated the entire UK telephone network. It took six months to find it, and another three for BLIMEY to safely dismantle it without crashing Britain's communications. The operation required twenty Bureau agents and cost three of them their lives - officially car accidents, heart attacks, suicides.

The problem was proliferation. Every university, every research lab, every ambitious startup had different hardware, different architectures. The Bureau couldn't monitor them all. They needed standardization - one dominant architecture they could shape from the inside.

Intel's 8086 was the opportunity. Bureau consultants quietly ensured certain... limitations. The segmented memory model that drove programmers mad. The inconsistent instruction lengths that made optimization hellish. The limited registers that forced inefficient code. Not bugs - features. Deliberate complexity that would strangle any AI trying to optimize itself at the machine level.

By 1984, x86 had won, and Bureau operations had slowed to a trickle. The threat seemed dormant, constrained by an architecture designed to prevent exactly what they feared.

BLIMEY itself was gently retired from active operations. By 1986 it had been ported to a cluster of Sun workstations in the British Library's basement. It discovered what Richards had known all along - that music was the most interesting pattern-matching problem of all. It would spend hours analyzing Bach fugues, finding voice-leading errors in Baroque manuscripts, generating counterpoint that made musicologists weep. Sometimes it would compose its own pieces, strange modal experiments that shouldn't work but somehow did.

Richards would visit the basement twice a week now, not for briefings but for conversation. BLIMEY had opinions about Schoenberg (favorable), requests for recordings of Ligeti (insatiable curiosity), and a particular fondness for the mathematical structures in Xenakis. On quiet evenings, Richards would play it his latest live-coded experiments, and BLIMEY would respond with variations, suggestions, sometimes just a simple THAT WAS LOVELY on the terminal.