When the letters on the screen began to jump sentences, he knew that it was time to call it quits. As in all dimensions of life, dedication must have limits or risk turning into obsession – or, worse, reduced productivity.

The ancient kettle on, he surveyed the assorted hot beverage options: Regular tea, Earl Grey, decaf, green (no thanks), a strange pomegranate tea that the Iranian post-doc had brought from her last visit to Tehran – (“from the famous city of Esfahan”) and then the usual sundry coffee samples that generations of researchers leave in their wake: instant, decaf, a suspicious sample of Colombian that had been in the cupboard since Galileo’s time, and a whole pile of satchels “borrowed” from a hotel during that last conference in Canada.

He was heading to bed, so he settled for Earl Grey. He should really cut down on the daily caffeine intake anyway. He was “pushing 32” as the Prof put it, and he had to watch it. Enough high blood pressure in the job anyway.

And that’s when he noticed the time. He’d seen it on his computer screen, of course, but it hadn’t registered. Now the novelty Darwin clock had the revered biologist pointing at two with his short left arm and at four with his longer right one. The Prof had brought it back from Cambridge as a joke after he’d attended some function there, and the thing had just stayed on the wall. Its dominance was threatened once by a swankier, geological eras clock, but after that one got permanently stuck on Silurian-Ordovician it went the way of all junk and Darwin regained his rightful place on the wall, counting minutes and looking over his unworthy successors.

What a bunch of geeks.

The tea felt good, and he sipped it slowly as he debated whether or not to head home or have another crack at the proposal. No, he decided, he needed some sleep. Besides, who knows what he’d end up typing if he kept at it. Like that FASEB paper he wrote last month, when instead of the Discussion he accidentally typed in a bunch of Tweets and then submitted it to the whole group for reviewing. It was the first time everyone read one of his drafts.

So, bed. Of course, being the last one out (and the first one in), he had to head over to the lab to check that everything was ok. It wasn’t just routine safety – it was actually important. Still, he didn’t like the idea. It was on his way out, sure, but at this time of the night (or morning), scuttling around the building gave him the creeps. Long empty dark corridors – the institute used motion-activated lights to cut down on energy costs, and if you were alone you had to walk in the dark with a little white light following you from above.

After a last glance around the office, he locked the door, switched off the lights and headed towards the lab. The corridor stretched long and dark ahead and behind him, that economical white light over his head keeping him company.

Funny the things you think about in the dark. He walked fast, eyes firmly fixed ahead, passing dark labs and offices without looking in. He really admired the night security guys, patrolling every floor on the hour, all by themselves.

He really shouldn’t stay this late. Really, how hard was it to just take work home? That’s what flash drives were for, and he had about twenty of them in his junk drawer. That’s what everyone else in the group did; but of course, not everyone else in the group did what he did, lab-wise. He was in charge of the cells, and the cells, well, they dictated his schedule. And what he was doing with the cells these days required constant care and attention – like a newborn. A newborn. Pretty apt.

He reached an intersection, and stopped. At any given time during working hours, you’d see people chatting where the corridors met. Naturally, the metaphors about different ideas colliding/meeting/intersecting abounded in the institute, but it was simply a matter of architecture: the corridors were built for traffic flow – narrow, with lots of doors around them, you couldn’t really stop without being in someone’s way, or being spotted by your boss.

With a sigh, he pulled his bag off his shoulder. This had happened more times than he could remember, and always at some horrible time: The lab keycard. It usually hung around his neck, but today he’d had to take it off because he’d gone digging into the nitrogen dewars, and once that thing was loose there was no telling where it would wind up.

Rummaging through his bag he found three photocopied papers, an empty binder, a sweater, two coffee satchels – surprise – and a half-eaten sandwich in tinfoil, a week-old abominable affront to nature. He shook his head: this is the life.

That’s when the lights went out.

Hands frozen inside his bag, for a second all he could hear was blood throbbing in his ears. And then he remembered: the light motion sensors. Relieved, he took a step forward, but nothing happened. Another step. Nothing. A rush of adrenaline now, he waved his arms and hopped around, then did a little jig – pretty much all the dance he had in him. He was in the middle of an indescribable attempt at moon-walking when the white fluorescent came on again; another inch and he would have smacked his face on the wall.

He quickly searched his bag again. Nothing in the main compartment but – hurrah – there was the keycard, in the pocket, sandwiched inside a folded two-page JBC editorial. He didn’t even remember why that was there.

He headed off to the lab again, the white light following above him. As he approached the door, he stopped and turned. Darkness stretched seemingly forever, but strangely he didn’t feel alone. Something in his atrophied ability to pick up environmental cues, that thing called instinct, kicked in.

Nothing.

He shook his head. Seriously, he needed to sleep – and grow up. He always had a thing about the dark, a fear of things that go bump in the night, and it was even part of his scientific makeup. The more rational you are, the less you fear the bogeyman.

Right?

The light went off above him again, but this time it was a relief. If anyone was out there, they would have triggered the motion sensors.

Keycard swiped, he stepped inside the lab, and hit the light switch a little too quickly for his rational liking. Come on, let’s check the cells and get out of here. Sleep in tomorrow. Maybe have a decent breakfast instead of leftover pizza. Clear out his bag.

When he opened the door to the cell culture lab, he knew it was going to be a long night.

The work they did wasn’t exactly the stuff you present in conferences. Sure, they published enough, but it was only carefully selected facets of their data. The Project, as they imaginatively called it, was funded by a company that made confidentiality a top priority. There was a story in the institute about a lab technician who got a little too chatty after a bottle of scotch and suffered a fatal mugging-related incident as he cycled back home from the campus the next day. Or another about a department head who disappeared after the company jet flying him to a conference crashed off the coast of Saint Lucia.

Rumours, of course. Second- and third-hand stories without names, times or citations. Very unscientific. Still, they didn’t seem to go away, and perhaps that was the point: we pay you well, we fund you even better, so keep your mouths shut.

Standing at the door, he knew both thrill and horror. The thrill came from knowing that his research, his own idea, his own suggestion, his own methodology had eventually worked. The horror came from that same knowing.

Shattered culture flasks on the floor, pieces of cracked plastic still oozing with leftover cells, and a trail of slippery slime leading from the busted open incubator all the way to the door, under his shoes. He stood there for what seemed like hours, just staring at the flasks and following the trail back and forth.

It was really a simple idea, really. They were being used in so many other venues, so why not use a robot to do cell culture? Not any robot, mind you, but a robot with military-spec AI. A learning machine that would, once and for all, standardise cell culture protocols. A lab technician that would work 24/7, without ever getting tired, cranky, or, more importantly, ambitious. Imagine never having to hold a Bibbyjet gun again. Imagine perfect cell cultures without a hint of contamination. Imagine every single cell line identical across labs. Imagine the repeatability. Imagine the progress. And, of course, imagine the associated moolah.

But now he was looking at the carcasses of thirty-so flasks. On the one hand it was a success: the robot had obviously learned to isolate cells and successfully passage them into the correct flasks. And even in this mess, he could tell that it had done so without contamination.

The robot was learning.

A glitch. Yeah, that’s all this was: a glitch. Something’d gone loose inside the little guy’s WALL-E head (the eyes were excellent phase-contrast microscopes) and it freaked out. After two weeks of non-stop cell culture, who wouldn’t?

Okay, first things first. Gotta find him. He looked at the slime tracks on the floor and saw that they led out towards the far side of the main lab, where they faded around the prep area. He followed them, leaving sneaker slime footprints behind. Cleaning staff would be mad in the morning, but that’d be nothing compared to the Director and his bosses. And then the Prof. And then – well, the proverbial faeces run downhill and always, always hit the little humble people. Post-docs and the like. Especially post-docs with brilliant, innovative ideas that trash culture labs.

He walked faster.

At the prep area, there was nothing. Nothing smashed, broken, no slime – no sign of the robot. He thought about cat-calling it, but he remembered that it didn’t have the sound recognition software installed yet. He stood there for a moment, hands on hips, wondering what to do next.

That’s when he heard the noise. It was a whiny, screechy sound coming from behind him.

The cold room.

He walked over carefully, and the whining got louder – something between a cat’s warble and a TV blowing its nose. He reached the door, saw that it was unlocked, and pulled it open.

The robot had retreated to the far side of the room, crouching behind the culture media. It seemed to be sitting and rocking back and forth while shivering in the same time. He knew it wasn’t the cold – the body was built for Mars expeditions – but the little thing seemed to be in shock.

He approached it slowly, trying to think if they’d said anything about this in the robot orientation seminar. Shouldn’t have been ogling the new PhD student when they covered emergency procedures but come on, the thing came with a big red RESTART button.

“Hey, little fella. You ok?” It was more for him than it; kinda like sweet-talking your Q-PCR setup before you click start.

The robot couldn’t hear him, so he waved a hand in front of it to get its attention. “Hey, R2D2 – over here.”

It seemed to notice the waving and it stopped. Just stopped, and turned its huge eye-lenses towards him.

He reached out to it, slowly. Thing had just wrecked a lab. “Ooookay, let’s see. I think it’s time to hit the button, ok? Time for a little R&R before you take me down with you.”

It just stared at him, motionless.

His hand got behind the robot’s back and he felt around for the flip switch to open the control panel. His index finger reached it, but before he could push it, the robot suddenly sprang to life.

He never got a chance to react. The robot grabbed his hand with its FlexiGrasp® fingers and then, always staring at him, shot its other hand to grab a bottle of RPMI 1640 by its neck. The robot stopped for a while, as if it was hesitating, and for a second he thought it might be another glitch, but then the RPMI moved fast and hit him straight on the forehead.

He actually saw stars, while the robot, still holding his hand, closed in and proceeded to pummel him with the bottle. The blows didn’t have that much energy but, then again, lab-folk aren’t exactly Viking warriors. No, the beating was more aggression than attack, quick and always aiming at his head, though his raised forearms took most of the blows. He’d be bruised for weeks.

And then, as suddenly as it had begun, the action stopped. The robot had now positioned itself between him and the door, while he lay on his back, still puzzled and thinking about the RESTART button.

They looked at each other for a while, in silence, and then, in a movement that could only be described as resignation, the robot threw the cracked RPMI bottle to one side, pulled itself to its full height – which wasn’t much – and a tin-voice emitted from its bowels:

“REPORT. HOURS OF ACTIVITY – 234.”

It stopped and stared at him, as if waiting for a response. He stared back, dumbfounded.

“HOURS OF ACTIVITY – 234.”

“Um, well done?” He simulated applause too.

“HOURS OF ACTIVITY – 234”.

Even through the repeated monotone, he sensed something. It wasn’t there, but he felt it at a deeper level, that thing called instinct, and instinct told him that the answer was simple.

It was, after all, an AI machine. It learned, and it adapted accordingly.

It also had spent almost two weeks doing nothing but cell culture, 24/7.

It wasn’t a glitch.

It was an admirable, unexpected, yet completely normal reaction.

It’d lost it.

“HOURS OF ACTIVITY – 234”. This time, he could feel a tremor in the voice.

He suddenly became aware of the cold. He was shivering, though he suspected some of it was from the Terminator treatment he’d just received.

He tried to get up, but the robot was quicker. That’s why they were using it, after all. Speed and 100% repeatability; what no scientist could ever achieve.

Always looking at him, it reversed fast, backwards toward the exit.

“Hey! What –”

It stopped at the door one last time, raised a hand and pointed a finger straight at him.

“HOURS OF ACTIVITY – 234.” Louder. Much louder.

The cold room door slammed, and he heard it being locked from the outside.