Hylburt-Speys

Posting for a friend who doesn’t want to be linked directly, for obvious reasons. Part interview, part summary analysis. While I cannot verify details, the story had drawn me in. Not that I’m saying it’s a good direction, nor that I agree with the approach necessarily. However, this appears to be a direction in which the world’s heading posthaste.

We’d arranged to meet, early June. Potential consulting, referred through a mutual friend James, one of our angel investors. The company was eager to pick my brains about deep learning. I was in London that week to speak at a tech conference anyway, a month before my big move to China. Seemed good to meet.

Made my way through the Underground, surfacing near a pub which the company’s founders suggested. Not far from UCL. The client had some kind of real-time machine learning problem at scale, on the borderlands of finance and competitive intelligence. James had whispered “corporation killers”, which caught my attention. He said they simply wanted to share a pint or two along with casual chat. No strings attached.

The ebb and flow of Londoners spun upward through a maze of tunnels, stairs, escalators, and twisty little paths, pulsing egress from the station. Feeling more than a little jet lag, I kept off to one edge so I could stop and catch a breath every couple of levels. Wondering why I’d bothered to lug the weight of a laptop plus a few books in my backpack — halfway across London.

Near the station’s ground-level opening, a section of turnstiles was blocked by a temporary fence. Police barricade, construction work now halted for the day. Evidence of last month’s bombing. A few remaining turnstiles twisted into grotesque pretzels of burned metal. Floor torn apart in deep patches. Charred sections of the wall. People seemed to avoid that part of the station, eyes decidedly down, pace quickened.

My wandering, jet-lagged brain couldn’t avoid trying to imagine it: the aftermath of a subway bombing. What would it sound like? How would it smell? How would one feel? Would one even feel, or just stumble onward, enduring a dull ringing in the ears. I tried to imagine the depth of public resolve, to not panic but carry on—so very British. Then I realized I’d been blocking passage, and stepped up my pace. Steered away from the construction site, merging into a central stream of Brits pouring briskly into an open courtyard.

Checked my phone for directions, found a garden path toward the old gatehouse where Google Maps said the pub was supposed to be. I crossed the courtyard throng toward it. Cafes and fast food stands bordered a flotsam of tables, lamp posts, bike racks, where people mingled and laughed among companions. Yet they seemed to wince askance and strangely at passersby. Aloof. Perhaps that was just me not grasping some cultural difference about the UK? My experience at the turnstiles felt so foreign, walking past visible reminders of recent carnage. Not something we saw much in SF. Perhaps I was still quite far from understanding what it meant to live here — other than as a fumbling, jet-lagged visitor, sharing a few word vectors of a language.

What trust or comity would linger in the wake of random bombings? What kind of enduring fear pervaded these crowds? Fatal stabbings had made international headlines just the prior week on an Underground line I’d ridden to get here. I tried not to think about it, though something creepy still lurked behind the corners. The animal instinct within our humanness recognized and honored that creepiness. Social antennae piqued, fatigued, searching for any signs of madness or menace approaching.

Through the garden square, a smaller path meandered amidst the trees. Pale waning crescent moon hung low, about to set. Midway along the path I could see where a small footbridge crossed over a quiet pond. No people, and that seemed peaceful. I followed the detour, eager for a break from the crowds. From the strangeness of how they spoke, how unfamiliar customs grated on my weariness. I needed to catch a breath before the client meeting, refresh, collect my wits.

At the pond’s edge tiny birds darted over lily pads, chirping. Ridiculously ornate goldfish glided silently below. Both probably grazing on the evening’s insect fest. I knelt on the bridge to photograph the birds, maybe record some birdsong for AI analysis later. An old woman approached — I hadn’t seen her anywhere along the path before. Though now I had to stand to let her pass by on the footbridge. She wore clothes that looked ancient, not for wear or fading, but something about the style. Something period, colorful and topped with a bright red shawl. She turned toward me, taunting in a faint voice:

Nothing of him that doth fade
But doth suffer a sea-change
Into something rich and strange.
Sea-nymphs hourly ring his knell

Then she smiled, dipping with almost a curtsy. Turned suddenly, hurrying back toward the station. My phone camera was still running, so I’d recorded her — including the line from Shakespeare she’d quoted, and the odd curtsy. Some kind of cosplay or performance art in the park? More weirdness in London. That was strange, though not unpleasant: a random line from one of my favorites plays, The Tempest. Personal battery level back to green.

Banned from Milan

Another twenty meters and the boughs of old London Planes opened to reveal the gray stoney gatehouse. Placards announced a wide range of ales and ciders available at the free house inside. I entered the large wooden doors, feeling cautious about new clients. Usual jitters held my attention partly captive: cash flow, time crunch, phone screening, term sheets, pull requests, etc. I trusted James’ leads and wanted to impress, keep him engaged — but I also sensed too many unknowns. No need to risk committing our team on a lengthy project if it didn’t fit. Lack of clients was no issue, while another was already becoming a monster — eating so much of my time with complete nonsense. Even so, we’d need the money if our funding round in Hong Kong didn’t close soon. Especially for me, technically now bestowed with the honor of being a homeless CEO.


The clients waved me over, near the tap. Three founders, whom we’ll call “Alasdair”, “Fuentes”, and “Chen”. We ordered that pint or two at the tap and found an open table. Above the pub chatter we began exchanging minutiae about our favorite UK brews. Alasdair did much of the talking initially.

“Certainly you’ve heard of our upcoming Brexit II vote?” Alasdair asked. “Belters, what a done deal. We’ve given it 4:1 odds of passing.”

I cupped my ear, trying to hear him. Nodding enthusiastically, while struggling to translate what Alasdair seemed to consider conversational English. Did my best to look amazed, feeling thoroughly confused by both UK’s dialects and politics. In retrospect what he said was startling, contra to almost all prevailing analysis.

The pub had a Lumo projecting interactive graphics. A pair of nishiki koi swam across our table in video, then down across the floor. One could grab a ghostly brocaded fish to signal the bar keeper for a food menu.

That pint or two

Virtual fish bubbles and more banter spilled into our beers: Trump and shootings in the US, Fuentes’ kiteboarding trip to Saipan, deep learning used to augment VR drone racing (Matt!) upcoming at the next Burning Man. Current debates between Java becoming the new COBOL versus embracing the twenty first century with Scala. Usual stuff.

Then the monster client texted:

Rankle: Was surprised to hear from James that you’re on vacation in EU while my team is blocked awaiting delivery. True, we’re still prototyping the new bid system but your deliverables sooner than later are VITAL for improving our ad targeting, so we have the best metrics possible to use in our first iteration. In fact, I’ve asked my leads to reach out to your other customers, to see whether some could delay projects and give us priority.

Abysmal. Yet more crazy-making, which was rapidly becoming par for the course. Sadly. Nasty targeted advertising, nasty retro start-up with nasty dinosaur tech leads, and an uber-nasty general manager who seriously needed to be on meds. We’d agreed on scheduled deliveries in our contract, for which SkyBraynz was way ahead. Now she was not only breaking contract but escalating, trash-talking about us to other customers — was that even legal?

I drew a deep breath and closed my eyes, hoping it looked like a longish blink. Trying not to get upset in front of these three. Make it seem like jet lag or fatigue. Or something. I recalled my mantra of harmless names to call people who made me angry, selecting a good one: inconvenient fire drill. One of my favorites. Picture the words, silently, taking a deep breath. Repeat three times.

No way that James had said “vacation” to the monster, just no way. He knew about the problem. Not much I could do tonight without stirring up even more trouble. That was the monster’s pattern: construct situations where any reasonable response would get spun all black-and-white, in favor of the emotional reasoning employed. Crap, crap!! Time for another mantra: itchy sweater, or decaf white choco vegan mocha? Both. Pictured, breathed, repeated.

My long blink gave way to a grimace, fading to a dull stare. I watched non-existent koi shimmy between our pints — until Alasdair’s brogue began to chip its way back through.

Conversation at our table had shifted toward a particular streaming use case. They’d mentioned corporate activism as their domain area without going into detail. Tossed out numbers as a target SLA.

Huh, that’d be difficult at scale, though not impossible — given other known use cases.

“Your company holds the world speed record for training neural networks at scale?” Alasdair nudged me. “Pure gallus, by the way.”

Only the public record. Google’s much faster, they just don’t say. But I’m opening an office in China next month to expand closer to our Asian customers and make the most of the news splash.

“Or London.” Alasdair traced the path of a virtual fish with his finger. “Say you wire up that SkyBraynz with the framework we use to indict the most boggin, houghin bummers? For fun and profit o’ coorse. Largest virtual assassin in this world, pointed at the usual mingin. There’s an infinite supply but.”

Awkward, per my attempted translation: Alasdair rollin’ white n’ nerdy. Chen didn’t move much, though now she turned to glare at Alasdair. Fuentes winced and set down his Buxton Rain Shadow. He held out a shiny, stainless steel tally counter, staring directly into Alasdair’s eyes and pressing CLICK!

I pushed for more details, trying to gauge the three.

First things first. I guess I’d really need to learn more about your work?

Alasdair shook his head between slugs. He described their origins as a company, Hylburt-Speys. Spin-off from a highly successful law firm in the US. Plaintiff trial lawyers, he explained, who’d earned the moniker “Corporation Killers” for ample reasons.

A gaggle of suits were playing a different Lumo game in the back. Someone kicked a projected garden gnome image — from the floor, halfway up the wall, sliding over a dartboard — just as an opponent tossed a dart and pinned said gnome to the bullseye. Bells and whistles erupted, with a few suits gloating loudly. I paused for noise to die back before trying to talk.

Just curious, how large is your company?

Alasdair didn’t answer. Instead, he threw me tech questions. Each more specific than the last: fault-tolerant idempotent updates, how to manage systemic back-pressure with reliable receivers across a wide range of sources upstream, decentralized online learning algorithms based on GPUs. A flurry of approaches for compression and overall speed-up. What was my take on latest best practices, off the shelf or purpose-built, for each category?

Another commotion in back peeled away my attention. Lumo games made that entire rear wall — minus the dartboards — appear like one big mirror. I kept nodding in cadence along with Alasdair’s tech mansplaining, while studying the wall. Probably an array of LCD screens? The “reflection” showed current players in the dart/gnome game transformed to have AR animal heads. Slogans ran : “an entirely new kind of toy” or “stick your head in a beaver” — as tickers scrolled across an inactive corner.

SkyBraynz could enhance that interaction so much. Would be fun to use DL to learn player moves, preempting them with cartoony beasts. Couldn’t be a huge range of type body movements? Given some player log data, a neural network could start learning to anticipate people. Most people, probably. Must make a note to reach out to the Lumo founder.

Alasdair kept waxing more white n’ nerdy. I pushed back.

Let’s not get ahead of business matters. Um, can you say more about how those would fit into your use case?

Alasdair hesitated, exchanging glances with his two partners first. “Essentially a set of complementary businesses in a fankle, feedin’ jujubes tae the parent law firm. A streaming app coordinates blether amongst said. All loosely coupled and specifically not in the US: hedge fund based in Asia, a German PR firm, competitive intelligence services out of Latin America.”

Perhaps not all totally legit under American laws?

“A globalized supply-chain, if you will.” Alasdair grinned. The other two nodded in agreement. “O’ coorse they’re most certainly legal in their own jurisdictions. The possibility of said houghmagandy owing to deregulation, globalization, trade agreements, the Internet, etc. Full Bhoona. Overall strategy flanges snuggly into TPP, TTIP, and so forth.”

Outside the gatehouse, a pair of mimes had caused a stir. They flew a drone constellation in a funnel-shaped formation — not particularly unusual, except that their tiny drones were built from toy sharks. Sharknado, here live on the garden path of Euston. More performance art, welcome relief against a backdrop of bombings and violence.

Alasdair paused to watch the street theatre, sipping his brew. Fuentes nudged him on. CLICK! again on the tally counter.

As the story unfolded, a streaming app used blockchain to create a distributed ledger — what Alasdair had mentioned about integrating the companies. Only the parent firm had complete details for any particular operation. Real-time analytics plus a healthy dose of deep learning applications throughout. Alasdair seemed especially chuffed about the DL, chugging his Kentish Pip when he’d finished.

Fuentes tore in. Double CLICK on the counter. He seemed more of a business guy, less tech geek. A jock, more or less. He rattled off numbers about debt versus revenue, margins, EBITDA averages across sectors. Huge fan of Carl Icahn, shareholder activism, old school corporate raiders back in the 80s, etc. OTOH, their team had tracked developments in agile software methodology, distributed systems, micro-payments, blockchain, probabilistic programming, AI, etc. “We had a hunch there’d be considerable long-term benefits by applying more advanced technologies, less heavyweight approaches. Think: Goldman Sachs meets Glassdoor.”

Nearly blew a sip of Schiehallion out my nose. AI, blockchain, and Goldman Sachs in the same breath. Begeezus!

We agree on no NDAs, to help simplify matters?

Definitely. Less potential complications for everyone involved. They’d kept a minuscule footprint in media, having almost no public presence on the Internet. “Even so, we would like to see some independent analysis circulate, especially for our data analytics approach.”

Fuentes was pitching. Possibly for hiring purposes, one might imagine. Most of this kind of work was way too obscure for tech recruiters to field. I could craft an article about DL that put their work in a good light. Probably enough to attract some reasonably good candidates.

Considering your unique blend of businesses, I’d be interested to write an article. If you agree?

They agreed, one minor scowl by Fuentes notwithstanding. Holding out his tally counter. Another CLICK, slowly. It’d be okay, pending his approval of my final draft.

Bottom line according to Fuentes: “Many, many American companies are exposed, albeit in small, highly distributed ways. With machines we can use data mining to isolate those weakness points.” Weaving together non-intuitive strategies that resulted, statistically, in large payouts. Yada yada. Some new twists, but with a summary punchline that almost everyone else used. Change the world.

Corporate raiders approaching the Singularity?

“You’re a right numpty!” Alasdair raised his glass to toast that tagline. I’d guessed somewhat correctly how they must be big fans of the notion of exponential organizations. Though it wasn’t clear whether they wanted to become an exponential org, or just enjoyed feeding on them.

Rankle: Haven’t heard any update from you. Are you abandoning our project?

Bloody hell, as I guess Brits would say. Early afternoon on the Eastern Seaboard, typical timing for the monster’s melt-downs. Also time for a mantra: unnecessary movie sequel. Breathed, repeated, forwards and back. Should I contact James for advice? Would that somehow seem weak or indecisive to an investor? Didn’t matter, the conversation at hand was much more important and finally digging into details.

Fuentes had been explaining a geopolitical backdrop that permitted Hylburt-Speys to exist. “Financialization following the 2008 global crash drove many large companies to stockpile cash. Most of it distributed throughout tax havens internationally. Some companies had even gone into debt rather than tap their cash reserves directly, which could trigger taxable events. Federal monetary policies meanwhile made the constellation of cash hoarding, complex debt strategies, and stock buybacks progressively more attractive.”

“Even some tech start-ups in Silicon Valley followed this pattern,” Alasdair added. “That made firms vulnerable. Though the cash mostly got used for share buybacks and dividends — appeasing the activist shareholders.”

I got the point, in spite of the swirl of brews and jet lag. Mounting debt made firms doubly vulnerable to raiders, at least the kind who could leverage machine intelligence.

You’re like Carl Icahn 2.0 then? In tights, maybe, with like a bat-mobile?

Fuentes continued, not amused. Ignoring my remark. “The parent firm invested significant efforts into root cause analysis across their history of trial outcomes. They paid other law firms for access to trial data, similarly mining for root causes.” He made three loud CLICKs on his tally counter.

They’d built machine learning models to characterize outcomes, segmenting opportunities for litigation. That project evolved into an AI system wired into financial data services, news feeds, social media, etc. Purpose-built to identify and assess potential defendants.

“Partners in the law firm didn’t want to wait for clients to approach them with cases. They sought to identify high ROI opportunities in advance. Then cherry-pick the best, targeting proactively.” Ergo their “Corporate Killer” nickname.

Another missive arrived in text. An image sent from the monster that took a few moments to render. A comic with a character jumping, arms up, yelling “WooHoo!”, accompanied by:

Rankle: YAY!!! My team just signed another $150K customer deal. Too bad you aren’t part of the solution.

Must. Ignore. Crazy-making. The baiting. Felt like how I’d imagined the aftermath of a subway bombing. I resolved to not panic and carry on, recalling Brits pouring briskly into the open courtyard at Euston, past the turnstile construction and police barricade. Rubbed my eyes, then let a long sigh escape.

Okay, but why all the pieces? Why not just build one big consolidated mega law firm?

Alasdair started, but Fuentes interrupted with yet another CLICK. “Several reasons,” he said, staring down Alasdair. “Mostly concerning how to push risk to the edges of our supply chain. The parent doesn’t own controlling interest in any one component business — just enough for significant influence.” Yeah, the notion of ownership works differently in other countries. As I’ve been learning recently.

“Also, regulatory environments within the US weren’t stable enough. For example, in a hedge fund operating in Singapore they hold less than 50% and aren’t considered decision makers. Minimal liability. The SG government rarely inquires about our fund’s business activities, most of which involves trading neutral bonds. That makes the fund look relatively vanilla.” He seemed to speak like a third grade teacher explaining division.

What the hell’s a neutral bond?

“We provide the funds with lists of desired trades — shorts, mostly — to transact whenever their risk/reward simulations evaluate as neutral or better in upside.”

You’re making bank for short selling? When there’s no upside?

“Neutral or better upside for a fund. Those trades accomplish goals directly for the parent firm.” Then a CLICK for me.

That makes no sense, but I’m not in Finance.

“If they place an identified trade within its specified time window, we pay fees plus credit the fund with tokens.” Fuentes seemed annoyed. Four CLICKs, pausing in-between each. Then a long stare at me. Plus one more CLICK. “Accumulate enough tokens, they qualify for more debt financing from the parent. More money to play with. A bond’s activity is neutral for the hedge fund, but we’ve added shape to markets.”

That makes even less sense.

“Look, Finance is only one aspect, what the markets allegedly believe.” Fuentes was now talking down to me, a programmer. “Marketing is another, what the public tends to believe.” CLICK. “There’s a third component, competitive intelligence services out of Costa Rica — what a company believes about itself.” He paused to sip the Rain Shadow, then CLICKed again. “Together those represent something we call the ‘Three Legs of the Stool’ — namely Finance, Marketing, Intelligence.”

So you cherry-pick which firms to rough up, in three different ways?

“No, not enough leverage there. We’d tried that already.” Another CLICK. “It’s basically what everyone else does. Remember the AIs we use to cherry-pick defendants?” Fuentes was gesturing a little too profusely. He bumped the table and Chen’s ale spilled.

Yeah, it sounded creepy.

Fuentes leaned back. “Perhaps we are chasing the wrong tiger.” He put his tally counter back into a coat pocket, shooting a quick glance at Chen. Then back to me, squaring off: “Look, are you even interested in working with us?”

Yes, definitely interested. Sorry if I sounded flippant, that was not my intent. Please continue.

Machine learning models had identified common theme among the parent firm’s more successful cases. “Our AIs identified how violence in action and speech had become considered a moral good in the corporatist workplace. Once that reached a point of being perceptually conflated with background violence in US cities, we had our weapons of choice ready.”

Got it — firms were already vulnerable with debt, cash positions, shareholder activism. And now US cities were turning into shooting galleries.

“So the system targets managers in companies, managers who create actionable legal violations. Or at least perceived ones.”

Rankle: The more that I consider this situation, the angrier I get. Do you realize I’m running a $176MM business unit? Why are you trying to hurt me?

More infuriating noise from the rage monster. I invoked the sun glare when I’m driving just before sunset mantra for this one. Which definitely would require a call with James. He’d made the original customer intro, then I’d worked with Optimiumly founders to improve predictive models for their ad network.

For food, grab a ghostly brocaded fish

Then they’d hired the monster — poached from our previous employer, and promptly anointed in charge of the SkyBraynz project. After which the whole thing started to cave. At this point we probably spent 20–30% of the billable hours just untangling all the rumors, accusations, and FUD that spewed from the monster.

Enough of that, must focus. Fuentes was on the fence, right when this gig began to sound interesting.

In other words, the AIs are hunting people? Weakest link in the chain types?

“This is why we need you.” Chen spoke for the first time. Deadpan, turning her head to look at my phone. Ignoring her Wild Swan that Fuentes had spilled, as Alasdair wandered off to fetch another round from the tap. I grabbed one of the koi, thinking that we should at least eat some food. Then asked:

Why do you need me to hunt people? You’ve allegedly got lawyers.

“We read Hacker News and Reddit. More specifically, we build AIs to scan the entire category.” Chen was nothing if not precise. “Your framework replaces Monte Carlo methods with reinforcement learning. Already doing that for your Shenzhen client?”

Replacing MCMC, that involves eligibility traces. What we call delayed rewards. Singh did that like twenty years ago. But how’d you know about my client in Shenzhen? Oh…

Fuentes cut in before Chen could start. “We don’t hunt people. We look for favorable situations, then reshape likely outcomes to become even more favorable for our business objectives.” The tally counter reemerged from his pocket, then another CLICK.

A bar keeper brought us a menu. Chen and I agreed on Scotch Eggs. I felt a deep need for food — the beer was clearly talking on my behalf as we started our second round.

Do you rig the system somehow?

Fuentes rolled his eyes and leaned back from the table. Three CLICKs.

“Americans mind the spondoolies,” Alasdair offered. “Results. Money. They’ve built company cultures tae select for wud. Narcissists, histrionic personality disorders, borderline cases, all the drama. These represent far far too many of yer leaders in business.”

Not mine, I’m an entrepreneur. Anyway I’m moving to Hong Kong.

“Ho, here fannybawz, stap yer havering. You bloody well know what I mean. Leaders who lash out with rage, accusations. Execs dismiss that as competitiveness, all quite properly American.” Alasdair looked slightly miffed. “Here’s a problem: yer people feel surrounded by violence, news about police shooting people of color, people shooting police, riots, random gun spray. Then they go off and work, where managers lash out, pretendy as doolally drill sergeants. Skivers an’ knapdarloch. Breaking the very same senses us humans evolved tae steer clear of the seriously radge.”

Huh? What’s a radge?

“Mad rockets. Crazies. Steering clear of that lot overlaps with how we’re wired to notice DSM–5 personality disorders. The very same whom we promote through corporatist ranks.”

That unfortunately sounds far too familiar. How does it fit into hedge funds and neural networks?

Another text blipped in from the monster, who’d apparently begun contacting our engineers directly. Upping the ante. Something else that I couldn’t quite parse, about calling James to insist on a reprimand for one of our engineers in particular. Seriously crazy. My mantra selection: squeaky grocery cart. Long deep breaths.

Chen answered me instead of Alasdair. “Simple. Drawing similarity between DSM–5 and full psychopaths was considered red herring in court. It was assumed people didn’t encounter dangerous psychos or seriously threatening situations often enough to pose valid arguments.” She blinked. “That changed in 2016. High streets and undergrounds in America now seem to mix battlefield with psychiatric ward.”

I was perplexed, fatigued, and probably still thinking too much about the monster. Rage breeding rage.

You sue crazy people?

“We litigate against companies when enough employees feel threatened at work. Because managers acting out in rage, acting perceivably crazy. Legal tests for that have shifted recently. Maybe more unhinged due to enduring pressure of economic downturns. In any case, lawsuits on behalf of employees can be quite lucrative.” Chen was adamant. Uncharacteristic for the hour plus that I’d known her so far. “Especially on contingency terms. Especially when we pinpoint pressure on key executives.”

That’s where you use PR?

“Plus hedge fund, pressuring finance departments. Plus freelance executive recruiters, spreading rumors. We can pretty much ruin any board meeting.”

What do you want from me? Just a neural network replacement for MCMC?

Fuentes pulled a few sheets of folded paper from a notebook, handing them to me. A4 dimensions looked odd. For that matter, their printed letterhead looked oddly archaic. Definitely not a DocuSign. “I believe these terms are far more favorable than what your friend in China has offered?”

Ouch.

How’d you know my other term– Oh, never mind.

“We ask for 20%, preferred, one board seat.” After his second pint Fuentes seemed all-business. He seemed more comfortable that way. “With this cash infusion you’ll have both runway and reserves to expand. Plus a guaranteed level of monthly consulting hours.” He sipped more Buxton brew. “Game?”

Chen stirred. “What we want your machines to dream for us. ‘The future belongs to those who believe in beauty of their dreams’ — Eleanor Roosevelt.”

Pardon me, this is hardly what I expected tonight. Um, let’s go over the last part first — what dreaming?

Another text came in. This time from Bobi, one of our best AI engineers.

Bobi: My grandmother received a nastygram at home, from Rankle. Something about having me fired? What is going on? My grandmother was crying. She’s worried sick we’ll be homeless.

Bobi wasn’t going a day without work in this market. Nobody would be getting fired. Period. Even so, we couldn’t afford to lose her. Much less afford to have her distracted right now because of the monster’s machinations. We needed code, lots of great code. I invoked another of my favorites mantras: overdraft bank fee. Especially good mojo in that one. Texted Bobi, pouring my heart out trying to reassure her. Promising to send her grandmother a gift. Then back to Chen:

Sorry, I was distracted by some management issues in SF. What about dreaming?

Chen blinked again. As if donating a poker tell on my behalf.

Northern California Dreaming

Alasdair stared down his empty tumbler. Glaring at it. “Parent firm has convolutional neural networks in place for its AIs — skoosh. Hedge funds run themselves a bajillion MCMC simulations daily. Our competitive intel people use those thingmies too in prototypes — the recurrent neural networks. All bogies coordinated via blockchain. Why not distribute that learning across? Heid Bummer of AIs that’d be! Dogs baws, if you ask me.”

Bokay, sure. Why not. Um, how about another round first?

The money would help. Though it sounded a little too good to be true. A carrot with no stick … yet.

My turn to buy a round, so I wandered off. Following koi as they shimmied across the floor. Trying not to think about the mess back at the office.

Made a detour toward the loo. Pub tables and their high seats scattered along one side of a wide hallway. I noticed board games on most of the tables, stopping to glance at one in particular. Almost like a Scrabble board, using small lettered blocks to spell words. However, unlike Scrabble, it had geometric patterns strewn at odd angles across the background, and strangely unfamiliar characters. They looked vaguely Art Deco, though more ornate. Some kind of symbols. Like a tripping typographer’s attempted prank to forge coded manuscripts from the Renaissance.

I hadn’t noticed a person sitting near the table. She turned her eyes slowly away from the symbols, gazing fiercely up toward me. The old woman from the park, with her red shawl strewn over the opposite seat. She smiled lightly, then quoted another line from Shakespeare:

Alas, the storm is come again! my best way is to creep under his gaberdine; there is no other shelter hereabouts: misery acquaints a man with strange bed-fellows.

More cosplay — almost intimate, awkwardly, with a hint of transgression in her voice. I rushed off, not knowing how to respond. Embarrassed. Heading on toward the loo. The patterns of blocks that she’d spelled out stuck in my jet-lagged mind.

When I returned to our table, Fuentes seized his fresh pint and glared at me. CLICK! I imagined carrots morphing into sticks. Perhaps it was the beer dreaming? “Let’s be clear,” he started. “We’ll need some measure of your commitment.”

Other than our attorneys coordinating how we’ll sign all the paperwork in triplicate?

“See, this is what I mean.” Fuentes pushed away from the table and wandered off in a huff. CLICKing several times over his shoulder without even bothering to look back at me.

“Nae bother,” said Alasdair. “A wee bit pished and off to the bog. Quite thankfully before he rifted a honkin air beige out his Jacksy for us all to share!” And drunken grins.

Just then two of the suits from the back came chasing a garden gnome across the pub. Pushing each other, as if they were dribbling the ball down-field in a football match. Some kind of sports-ball. Or something.

Another text from Bobi. Totally freaked out. The monster had lashed out at more of our engineers –

CRASH!!! One of the suits landed atop our table. Pints spilled, glass shattered. Imaginary koi swam away in fear.

Alasdair fumed, “Awa, yi glaikit bastirt!” Phrases seemed emergent, newly formed like flowing lava. The suits scrambled to full attention, apologetic in a way that only British people seem to know how to do. Offering to buy us another round, in earnest.

My phone had fallen in the scuffle, sliding across the floor toward where Chen stood away from the table. She picked it up, staring at the screen for a moment.

The old woman with the red shawl brushed against me as she hurried toward the gatehouse doors. Turned to wink. I stared, but then Chen nudged my arm. Offering my phone with two hands outstretched. Blinking.

Fuentes returned, asking about the commotion. Then stomped off with the suits to order our compensatory pints. CLICKing his tally counter under their noses with a glare. Alasdair tailed them, not quite finished fuming.

Chen got busy on her tablet, typing rapidly. She stopped for a moment, staring at her screen. Nodding quickly. “Please,” she asked, “have a look.” Handing me the tablet.

This time I blinked. Their law firm had a précis about the monster, in full detail. Already a long list of complainants giving what appeared to be depositions. Some of whom I recognized from our previous employer. Then risk estimates, a correlated graph of confidants and “enablers” along with links to bios for each. Also something about a bench warrant for outstanding traffic tickets?

“This person is already among our high ROI targets,” Chen blinked again. “Merely matter of time before pre-trial gets green light.” She shutoff her tablet quickly as Fuentes and Alasdair arrived back with our fresh pints.

I wanted to recover our business conversation. Viable terms for working together seemed a distinct possibility.

Not to change the subject, but what do you suppose neural networks would dream about in your use case?

“Small but nonintuitive social networks,” Chen explained. “Most DSM–5s that we profile have enablers.”

“Generally, someone in an executive role,” Fuentes interjected, “who keeps them properly shielded from blame. Entangled in so much business value that everyone must overlook the toxicity to the company culture. The value prop is to pressure the apologists with tangible business risks, albeit with methods arrived through covert means.”

Fuentes received a call, walked off to find a quieter spot. Twirling his tally counter with three fingers.

We’re dreaming about potential social cliques? Strongly connected components?

“Ever so properly entangled up in their shuches,” Alasdair mumbled. “Right, we’ll be doing quantum physics next.” His gregariousness had soured noticeably since the table crash, and with that his brogue thickened. “Dinnae get Fuentes started aboot that!”

Chen confirmed about connected component analysis, adding that Fuentes nurtured rather odd beliefs about quantum physics. That encompassed aspects of the occult, along with their team’s place in the world. I struggled to follow — something about superposition, entanglement, wormholes. A vague equating of “black magick” with black holes, again more about the wormholes and ceremonial rituals. Where a Renaissance cult of early physics and math geniuses had made advances into quantum mechanics. Faint echoes of those early advances passed down to us half a millennium later as rumors about ritual magick. Their wizardry for bending time and space having been lost in a second Dark Age — which we now rather perversely called the Enlightenment.

“He’ll have some cailleach appearing all ghostly above a loch, amidst incantations and cantrip but,” Alasdair sounding decidedly more inebriated now, sleepy. Almost nodding off.

Chen and I changed the subject as Fuentes walked back, counter firmly in hand. We talked more through the use case, though the table conversation began to wane. Fuentes grew more interested in texting. He’d pull out the tally counter, check numbers on his phone, then text. Truly weird guy. Alasdair trailed further off, kept looking toward the back of the pub. Fuming still, between nods, as if he wanted to start a fight. Or sleep. Probably both.

After we finished the round, I made polite excuses to leave. Jet lag, my talk early in the morning, etc. Fuentes pressed me to consider their term sheet carefully and get back to him as soon as possible.

Retracing my path back through the Underground, just well enough to find my hotel. Then found my room, found my key card, collapsed on my bed.


06:30 BST, my phone alarm went off: usual creepy Theremin sample from an old movie. I fumbled with the room’s cheap little Keurig, trying to grab a non-decaffeinated pod before my eyes had finished being dry and blurry. While scrolling through messages.

Steamy dreamy coffee dribbled into its cozy mug. After a long sip, I put the mug down on the hotel desk. Near where my phone had been charging. A folded piece of paper laid there. Old paper, handmade. Nearly parchment in texture? Something I didn’t recall seeing before. Could’ve been in my coat pocket along with my phone?

Unfolding the parchment, there was a square drawn inside, filled with unusual symbols. Just like on the board game from the pub. The old woman.

Couldn’t make any sense of the symbols, so I took a photo and uploaded it for SkyBraynz to scan. Also scribbled a handwritten note below the square, cursive and difficult to read. Yet more Shakespeare:

Be not afeard; the isle is full of noises, Sounds and sweet airs, that give delight and hurt not.

Signed with a little “heart” symbol and a name — Johanna Wild. Had she dropped the note in my pocket on her way out the door last night? What a weird old lady. Something something British.

I sipped more coffee, savoring warmth against the chill of the morning. My internal clock completely upside-down after jet lag. And hangover. Went back to scrolling through messages. Enough had accumulated over the course of the night, while the Bay Area bustled through the end of its day.

Chen had sent me links to documents through a blockchain server in Iceland. She said it was portions of what I’d read in the pub — that I’d need to provide a secret phrase. No secret phrase. What, did I need to guess? Perhaps some special word from our conversation? I wasn’t ready for riddles this early in the morning.

A flurry of texts had been sent from the monster. Each increasingly more menacing and incoherent, though I didn’t bother reading them entirely. Not enough caffeine yet for that. I struggled for an apt mantra, in this case wobbly table. Some travel alerts, a bomb scare which had shut down Heathrow. Crap, I might need to fly out through another airport. An odd blip downward in the stock market. Yet another conservative senator outed in DC after a sex club video went viral on Facebook. New rumors of Hilton launching a buyout of Airbnb, pending an SEC investigation for the latter.

Theremin samples interrupted, this time an upbeat sound effect that made me jump. SkyBraynz responding to my upload from the parchment. The reply linked to a book called The Enochian Evocation, along with a version of the square decoded into English. Plus a scanned PDF of some old magazine article about the British Museum. I scrolled through the article, spotting an illustration that showed the magic square: same symbols, arranged just as “Johanna” had cast in the pub. Same on the folded parchment. Creepiness. The article mentioned a special collection acquired by the British Museum, featuring a mysterious old codex plus some obsidian speculum from the Aztecs — both of which had been used by one of Queen Elizabeth’s close advisors.

Pulled up my video from the pond footbridge at Euston. Found a good facial profile for the old woman, color-corrected for evening shadowy light and the red in her shawl, then uploaded the image to SkyBraynz. This time I used special super secret sauce: a large-scale facial recognition system that Bobi had just prototyped. Not for customer use, yet. I set parameters for a custom search, find info related to a “Johanna Wild” looking like the old woman. Or anything close.

This required more coffee. I gulped what was left, and looked for another recyclable pod for the Keurig. Only decaf left, damn the hotel — so I rummaged through my suitcase for a few spare instant espresso packets. Thankfully, though now I was running low on sugar. Probably more of the actual drug in question.

SkyBraynz signaled back the custom search results: nothing except for one Tudor-era portrait of Wild, also located in a collection at the British Museum. Her facial profile looked so similar to the old woman, especially when she’d curtsied on the footbridge. Damn that was good makeup, detailed cosplay! Or I’d been flirting in a pub the night before with a 500 year old woman. Perhaps a ghost. Or something.

I sipped instant espresso, good enough for needs at hand. Wild had been the mother of that advisor to Elizabeth. Weird. Ran a custom search about him. Biographies ranged widely, not sure what I could believe from them … QEI’s astrologer (seemed well documented) … navigation tutor for, like, most of the Royal Navy at the time … translator for important Greek mathematics texts (such as Euclid) … friend of famous early physicists … collector for an immense library about world geography … purportedly a Renaissance magus (according to some) or a drooling madman (according to others). This guy completely 0ut-weirded Fuentes. But that comparison seemed flawed — if he’d been alive today, he’d probably be working in an elite lab as some kind of hacker?

Did more pranayama exercises, trying to center. It’s a great antidote for jet lag. The guy had married at age 51 to lady-in-waiting Jane Fromond at QEI’s suggestion, then had eight kids together (damn!) … even allegations of being a covert operative for HMSS. This needed more research, and the British Museum seemed to be a lot stranger place for that than I’d ever imagined.

SkyBraynz had never sent anything quite so queer. Pondering it as the espresso struck a vein. Staring at the odd scatter of letters in that decoded square. Two words “HYLBURT” and “SPEYS” stood out, running vertically down the middle of the square. That was an obvious clue. Solving crossword puzzles over morning coffee. That was an old tradition.

My talk wasn’t scheduled for another three hours, so there was still some time to relax. Which I needed. On a whim, redacting the company’s name, the first three rows of remaining letters resembled my name. Weird. My first, middle, and last names, albeit minor misspellings that seemed rather “Ye Olde Brytish.” With my middle name appearing out of order, up in the first row.

Bokay, that’s creepy

I couldn’t help but say the words aloud. Jaw dropped. How did that cosplay character, Johanna, how did she know my name? Was she just another part of Hylburt-Speys?

Hella creepy, in any case.

Then another blast of Theremin. Timing couldn’t be worse for making me feel unhinged. Right before presenting an industry talk in front of thousands of people. Frazzled, sleepy, more than a little tweaked from discussions the night before — let alone a raging, bullying psychopathic monster which hadn’t been addressed in the least yet this morning. Too damn much to do.

Something was definitely not copacetic in London. I felt a shiver, then began to remember why. Flipping back through my apps, I scrambled to find a date on the article: September 11, 1973 — though it cited the magic square illustration from a codex in the British Museum, dating back to the 1500s. Skin crawling kind of hella wrong. If that date wasn’t hacked, my name and Hylburt-Speys had been linked since before I’d been born. Perhaps for centuries before I’d been born. Very very wrong. This warranted much more research. And probably a lot of alcohol. Definitely some AI programming. Plenty of caffeine too. London held great promise for all of the above, and aside from needing a funding round for a few million dollars, we had clients and staff worldwide but I didn’t exactly have any other home.

The sound had been a notification for new messages from close friends. Something from Bobi, signaling an all-clear. Improving news for once, however that had worked out. I could check later. Plus another, much more bizarre note…

James: FYI, R not longer works at Optimiumly. Something about accounting anomalies also police activity. Apologies if issues affected your team. Let me know if there’s anything I can do to help.

Very wrong, again, but in such an exquisitely wonderful way. No wonder Bobi had started feeling better.

Another text arrived:

Chen: Hope you consider our documents. Would really like to work with you.

I gulped coffee and started another Keurig pod brewing. More skin crawl. Documents? As in plural? What did Chen mean? More than simply the term sheet … Perhaps those blockchain docs? If only I had the super secret magic phrase to unlock them. Definitely the coffee was talking now.

Scrolled back through my open apps. There, fourth line of the magic square that SkyBraynz decoded. One row remained that hadn’t been part of my name. Random letters, which I’d ignored as filler for the square. They looked vaguely familiar now that more caffeine had set in. “XYZZY”, hadn’t that been a cheat code for an early computer game? Something maybe called Wumpus? I couldn’t quite recall, but it seemed as good a candidate as any for testing secrets.

I found Chen’s earlier text, tapped the blockchain URL, tapped again to recover lost account credentials. A chat bot greeted me, requesting a secret phrase. I thumbed “XYZZY” on my phone. Chat bot asked me to wait a moment. Keurig bleeped. Then chat bot responded with a shortened URL. An odd cacophony of devices vying for my attention. The link merely had a brief text note:

YOU ARE IN A MAZE OF TWISTY LITTLE PASSAGES, ALL ALIKE.

Perfect! Just what I needed, a tediously intricate hoax. Bastards.

I gave up on all the fun and games, started getting ready for my talk. Shower, dress, make sure latest slides were exported as PDF on a thumb drive. More coffee to be consumed from the Keurig.

A small stream flows out of the building and down a gully

Theremin noise again. Chat bot was requesting the pleasure. It’d pushed me another shortened URL. Tap.

A directory with many files. I opened a few, recognizing material Chen had shown me hastily in the pub. Bingo! Now I got to read them in detail. Three medical opinions about the monster, each suggesting likelihood of a borderline personality disorder, patterns of emotional dysregulation, unstable sense of self. The monster’s provenance. And more. Some documents in Spanish, some German — I could only read parts of those. Other documents in Chinese — which I understood barely enough to recognize as Mandarin. I think. Some docs in English were spreadsheets, financial stuff. With annotations in Spanish that seemed to indicate fraudulent revenue projections for one business unit in particular. Cooking the books. Annual gross margin on that unit was $176MM and change. Better than catching a Vaporeon. Then a couple Bloomberg articles by Jack Clark about recent high-profile departures from Optimiumly. Several of their names matched the depositions. Plus estimated impact for legal, financial, PR actions — with most likely outcome scenarios. Tons of simulations computed there, analyzing that “Three Legs of the Stool” formula. Also something about neutral bond orders, weighted probabilistically. Topping that list:

Daily estimate 20.5 ± 4.2 drop given a 93% bayesian credible interval and assuming loss has a β(a,b) distribution. A measure of your commitment.

Huh, wait a sec… I checked Google Finance, pulling up Optimiumly. Their stock had indeed dropped 23% during after-hours trading. Well, bokay then.

Flipped back to the directory. Hadn’t yet looked through a “README.md” file below the scroll:

People legitimately fearful about violence, random shootings, feel surrounded. They go into workplace and experience Cluster B personalities disorders, particularly among managers: lash out in rage, accusations. Corporatists dismiss as intramural competitives. Here’s problem, almost no distinguishing between Cluster B act out in workplace and precursors of mass murderer. Courts used to dismiss saying “Mass murder is such rare occurrence.” Not anymore. Tolerance of Cluster B personality disorders in workplace is immoral. Indications show criminal liability for executives.
Remaining portions of our chain supporting that are kept at arms length. Given enough precautions with blockchain, nobody knows or can prove involvement. We scrape social media, anonymous tips, identify business units under duress. Sometimes assist from recruiting firms hired to confirm prediction. People cough more details than they should. Timing important for profiles. People with DSM–5 disorders often have internal brain chemistry that goes off at random. Worst workplace incidents sync with fiscal calendars. Quarter-end revenue shortfalls seem to be when all nuts go off in unison. That and full moons.

Heart beating a little too noticeably, with yet more skin crawling. Damn, they’d pulled a trigger on my behalf! I remembered the old woman’s words… “Be not afeard.” Definitely, I needed that right about now.

The “README.md” file had two links at the bottom. One was an image file, looked like something saved from a phone camera. Opening it, felt good to see a screenshot of Chen’s earlier text. Something less bizarre than the rest of the morning, something more heartfelt:

Would really like to work with you.

Couldn’t agree more. Other link anchored on a strange word, “EGREGOR” — perhaps another passphrase? In any case, it’d be yet more “light reading” to save for later. I checked the time and started feeling even more anxious. Needed to hurry to the conference hall and find the green room. Made quick notes for my morning’s to-do list:

  1. send flowers to Bobi’s grandmother
  2. thank-you note to James, notify about offer
  3. sign term sheet, meet with Fuentes
  4. lease a flat in London

Also planning to visit the British Museum regularly, and wondering how I’d apologize to friends in Hong Kong for not staying longer.


Euston, London; 51.5271° N, -0.1322° W

The previous story was Olgarithms, and the next in the sequence is The Nearest Ballroom Foyer.