I emerged from a rabbit hole, blinking as normalcy returned.
Physically, I was running on fumes. A long series of late nights. Investing the hours.
For once, this wasn’t caused by the daily grind of wake, commute, work, commute, sleep. That gilded cage we construct for ourselves, when working to live degenerates into living to work. The infamous lifestyle trap, where nice things we own, and the nicer things we desire, begin to own us.
I hadn’t succumbed to peer pressure presentism, nor chosen to compete with careerist cutthroats.
Mid-life crisis clichés remained unindulged. No, the cause of my exhaustion wasn’t hookups. Drink. Drugs. Trophies. Wanting to feel alive. Desperate to feel at all. Wanting to be needed. Hoping to be wanted.
It wasn’t caused by an existentialist crisis. Realising that the point of “enough” is identified as the moment that the struggle for survival gives way to the struggle for meaning. Self-sufficiency swapped for self-indulgence. Curable only through a combination of reflection, acceptance, and surrender.
No, this time my flat batteries were entirely self-inflicted.
A spark of inspiration.
An idea embraced.
A leap headlong down the rabbit hole to explore it. The more I learned, the deeper I explored. The intellectual excitement of an Indiana Jones-style treasure hunt, with fewer snakes or Nazis.
I started out intrigued.
Soon caught myself feeling happy.
Contentment followed.
A feeling that sneaks up on the unwary. The welcome return of an old friend. It had been a while.
As my curiosity expanded, work started getting in the way. I didn’t want to waste time doing bullshit tasks or surviving the corporate Game of Thrones. I wanted to go deeper down the rabbit hole.
Exciting. Like the beginnings of a new romance.
A sick day here.
Intoxicating. As only a newfound addiction can be.
A couple of days of unplanned annual leave there.
Stealing time from what I was supposed to do. Making time for what I wanted to do.
Riding the wave of interest and motivation.
Old enough to understand it wouldn’t last.
Wise enough to recognise when to shut up and enjoy the ride, living those glorious fleeting exhilarating moments of freedom. Surfers and skiers would understand. Skydivers too.
The zeal to buy time forced me to apply some rigour often lacking, and overcome longheld reluctance.
I still had to attend. Be physically present. Go through the motions. Appear to play the game.
But that didn’t mean I had to sacrifice those precious few daily hours of attention and focus at the altar of paid employment, on priorities that were decided for me rather than by me.
When they were younger, my kids used to describe those attentive hours as having their “good brain” on, which once exhausted saw them “catching the stupids”. Beyond that point, it didn’t matter how many additional hours they spent grinding through schoolwork, the knowledge just would not stick.
Jeff Bezos made his billions through self-awareness. Understanding he was motivated by minimising regret, and good for at most three considered decisions per day. That doesn’t sound like much, but get the sequencing right and the compounding power of consistently making three correct decisions each day is enormous.
Warren Buffett’s investing success can be attributed to patience. He concedes that the vast majority of his fortune was generated by getting just one investment decision right every five years or so, then not blowing himself up in between times through boredom or the need to be doing something.
The trick in all three cases was making the space for productive use of that scarce concentration time. Recognising when it was, and when it wasn’t. Arranging and allocating activities around it. Ensuring busyness wasn’t conflated with effectiveness, in a similar manner to the myth of multitasking.
A trendy term from the buoyant job market of a year ago was “quiet quitting”. The act of mentally checking out from a job, and consciously deciding to phone it in. Doing the bare minimum to be retained rather than fired, but no longer striving for excellence or going the extra mile. Coasting through the work week. Participating, but not engaging. Present without being “present”.
I was raised better than that. The notion of not doing my best work offends me. Not from guilt or obligation, but rather because doing so would be letting myself down. If I was going to invest the time in doing something, why not try and do it well? Otherwise, I am cheating myself, so why bother at all?
I remember reading in an old management textbook about Martin Barnes’ quality triangle. It nicely illustrated quality being determined by the tension between cost, scope, and time. Unsurprisingly, my employer was uncompromising. Demanding the same volume of output at the same quality, no matter my personal preferences. I am after all just a not-so-humble cog in a corporate machine. Like all white-collar employees, I am infinitely replaceable at my employer’s whim. All they need is an excuse.
If wanted to free up my productive thinking time, I needed to work smarter.
To have survived as long as I had, the basics were already covered. Deliver. Add value. Be visible. Ignore noise. Make the bosses look good.
So what options were left?
I realised that I faced a variant of the classic economic question of productivity: how could I produce the same outputs from fewer inputs?
Thinking back over history there were many examples of this in action. The printing press. Steam engines. The telephone. Computers. The internet.
Each time it was a case of doing the same things, but faster. That speed was delivered by technology.
What technologies were available to me that could achieve the bonus seeker’s nirvana: do more with less?
One of my favourite genres of escapism is science fiction. Cyberpunk. Steampunk. Space opera. The more imaginative the better. A common device used throughout the genre is the concept of a virtual assistant. Providing the protagonist with the ability to offload tasks not requiring their full attention. Conveniently allowing the story to travel at pace by skipping over the boring hard work.
Fifteen years ago, Tim Ferriss made a small fortune teaching the world to buy back time via outsourcing the doing parts of their lives to lowest-cost assistants in Manilla, Medellín, or Mumbai. Democratising the long-established C-suite playbook of using executive assistants to do the scut work, keep them propped up, and make them look good.
It’s an approach that has some merit. In the professional world, I’ve long outsourced the production of slide decks, diagrams, and briefing papers. Hire a ghostwriter. Give them a brief, a deadline, and a handful of US dollars. Review the output the next day. Rarely is what gets produced the finished product, but giving something a quick polish is a lot quicker than penning it by hand.
Today, the media is full of breathless stories about the magic of machine learning. The promise of artificial intelligence. Large Language Models. Natural Language Models. Pattern recognition and extrapolation.
The end products remain a mystery, but the raw tools those products will be based upon are readily accessible to those who are willing to seek them out. But what use might they be to a white-collar professional looking to scale their output without increasing their effort or expenditure?
The premise of generative models is simple: most of what we do consists of simple repetitive tasks.
Consider the humble birthday card.
Write one birthday card, and the message might be considered heartfelt, thoughtful prose.
Write a dozen, and that same message might as well have been applied with a rubber stamp.
Repetitive
Clichéd.
Scale that across some 8 billion people writing multiple birthday cards every year, and those messages become a predictable production line of trite good wishes following a conventional pattern.
Once that pattern has been recognised, it can be automated. The automated greeting card message writer will become a multi-million dollar revenue stream for a company like Moonpig, which already handles the logistics of printing and dispatch of greeting cards on-demand.
That is the premise behind the hype.
The reality is that AI will be invisible, baked into the products we already use.
Now apply that same premise to the majority of what white collar professionals do to fill out our days.
Send and respond to emails.
Prepare business cases.
Provide status updates.
Manage budgets.
Review reports.
Write code.
Diagnose illness.
Every one of those tasks is repetitive. Rule-based. Following a convention. Paint by numbers.
Accounting standards.
Design patterns.
Industry best practice.
Legal precedent.
Standards of care.
Principle-based regulation.
Patterns. Patterns. Patterns.
If a model can discern those patterns, then an algorithm can generate or apply the correct solution.
Quickly.
Repeatably.
Consistently.
Maybe even correctly.
I’m not talking here about the laughably bad customer service chatbots rolled out by your bank or utility supplier during COVID so they could eliminate their call centres. Nor the shambolic Alexa/Siri/Google voice recognition abominations who promised so much and delivered so little.
No, I’m talking about being able to quickly generate credible professional quality outputs.
The sort of thing you or I might produce, when applying our cumulative skills, experience, and judgement. Firing off a quick email. Compiling a status update report. Designing a business process. Tuning an algorithm. Debugging some source code.
I decided my best shot at liberating my time and attention from my job was to apply ChatGPT. My employer had recently implemented a private instance of the model, to allow enterprising staff to make use of it without risking data leakage from their inputs being used as training data. The operating costs were trivial, less than the firm spends on toilet paper each month.
Instead of the cursory playing with the large language models I had done a few months ago, I went all in. Getting the robot to do everything from drafting emails to generating request for proposal documentation.
Where appropriate I would ask the model to be creative. Make recommendations. Explain and justify reasoning. Call out assumptions. Argue the other side of a debate, to explore all the salient points.
The results weren’t perfect. But they weren’t bad.
I learned to improve articulating clearly what I wanted. How to refine and revise what was produced.
By the end of the first week, I was getting through a day’s worth of output in less than an hour.
My goal of preserving my “good brain” hours for my nocturnal intellectual exploration achieved.
By the end of the second week, I was consciously having to hold back outputs that had been produced “too quickly”. If my turnaround time was too rapid, it would give the game away.
Work still felt like it got in the way, but now it was the inescapable time consumed by meetings, not by having to be intellectually “on” all the time. My stress levels were lower. My demeanour more relaxed.
This gave me pause. Moving to a working pattern of asking the robot to have first go at doing the thinking exposed how little genuine value-add I (or my colleagues) truly provide. My ego didn’t like that conclusion. Not liking something doesn’t make it any less true.
By the end of week three, I was comfortably coasting. Same scope. Same cost. Same quality. Only delivered faster and with less effort on my part.
My quest for productivity had been won. The protagonist and his trusty virtual assistant had slain the corporate dragon to rescue my good brain before it could catch the stupids by working long and hard.
Alas, all good things must come to an end, and so it was for my newfound intellectual arbitrage.
The corporate ethos at my employer is to keep piling on work until the employee is crushed by it.
After loading me up with a couple of additional projects, a suspicious C-suite occupant started to question why I hadn’t collapsed under what he considered to be an unreasonable and unsustainable workload?
My ChatGPT virtual assistant was rumbled.
The subsequent blast radius and collateral damage were as predictable as they were disheartening.
I was summonsed to attend an early morning “breakfast” meeting with the exec. So too were nearly a dozen of my colleagues, who each perform similar roles to me.
Conducted before the bulk of the workforce arrive at the office, “breakfast” meetings leave time to clean the blood off the carpets after somebody inevitably gets sacrificed, fired, or thrown under a bus.
That day was no different.
The exec praised my initiative and ingenuity for finding new ways to increase productivity.
He rounded on the rest of the group, berating them for not having already thought of doing the same.
With the performance bar raised, current headcount was no longer required to achieve desired outcomes.
Two of my colleagues were terminated on the spot.
One was a bit of a passenger, the slowest in our herd. The other had been living on borrowed time after a high profile project blew its budget and turned into a raging dumpster fire.
The exec offered the survivors croissants and coffee, as our former teammates were escorted from the premises by security. Since that day, I’ve been on the shit list with my remaining colleagues. Each forced to instantly adopt new approaches to plying their trade, as workloads substantially increased while pay remained the same.
Sink or swim.
Survival of the fittest.
The exercise proved useful. I learned some new skills and ways of working. I’ve also relearned the importance of prioritisation and decision-making. Make them for myself, or have them made for me. I still enjoy diving down the rabbit hole, though the novelty has started to wear off in a similar manner to the magic of a new relationship fading into routine.
At least I have my trusty robot virtual assistant to keep me company and help make well-informed decisions.
I must confess that, in a moment of weakness, I asked the robot how it rated the chances of my profession still being around by the time I reached retirement age. The response was diplomatically phrased and disclaimer-ridden, but could be paraphrased as: “good luck with that!”
Featured by
pathways 1 October 2023
Great post! I am intrigued about what you were doing in the free time you created.
A friend is currently using chatgpt at work, self-taught. He was reluctant to reveal this at first. But now it is all out in the open and yesterday gave a presentation on the benefits to the company and to the individual. I think this will be adopted by forward thinking companies. Those who are not prepared to give it a go may be passed over for promotion or loose their positions eventually.
I feel like industry has been through this before in different guises but this time it is an open source code and open to individuals to make the changes to industry.
No one has been fired at my friend’s company…they are already working flat out after the last round of redundancies.
{in·deed·a·bly} 1 October 2023 — Post author
Thanks pathways. It’s great to hear your friend was recognised for innovating their ways of working. The end result is the same, once magic becomes science then everyone is expected to use it. Call that evolution, or progress, or whatever.
Luddites will protest loudly for a time as they are left behind.
Everyone else will just re-skill and carry on.
In a year or two using these tools will feel as normal as Excel or satnav or cashpoint machines.
Not directly. But the next round of redundancies will be larger, enabled in part because not so many expensive bodies are required to produce the same level of output. It’ll be badged as “freeing up their time to focus on more productive activities“, however if the market doesn’t expand to support the sale of a greater number of widgets, then producing the current number of widgets at a lower cost is the next best way to improve profitability. The performance bonuses for those delivering the increased profitability are the same either way.
David Andrews 2 October 2023
Rather odd that your employer had directly facilitated the workload which they then considered to be unreasonable.
Sounds like you should have throttled your output a little more to avoid your stealthy assistant being detected.
{in·deed·a·bly} 2 October 2023 — Post author
Thanks David.
It is a flex, in the same vein as running fire drills or making people jump through hoops just because they can.
Some perceive they are deserving of being seen as benevolent by relenting on unreasonable asks, attracting the gratitude of their victim. A bit like Stockholm syndrome, the misguided loyalties of a battered spouse, or a whipped dog.
Alternatively, they may be genuinely trying to break or push out the minion in question. Sometimes intentionally, as it can be cheaper and easier than making them redundant.
Or they could simply be oblivious to it, which in my experience is usually the case. Self-absorbed and incompetent in equal measure. The same type of shortsighted exec that demands a rocket ship but only provides funding for a bicycle, or spoiled child who is forever adding items to their wish list without ever removing anything.
Over the years I’ve had many clients or bosses who fall into each of the three groups, occasionally more than one.
Seluron 1 October 2023
I wonder if you ever felt that the output produced by ChatGPT wasn’t “you”. Like, it is often polished stuff, but didn’t result from my own thought processes. It feels alien to me. I guess that’s fine for corporate busywork I am obliged to do but don’t care about. But for example by book writing I am emotionally invested in, and I just couldn’t get used to have the model write for me. I went back to the slower process of doing it myself, because the result felt “mine” and I am able to feel proud of it.
{in·deed·a·bly} 1 October 2023 — Post author
Thanks Seluron.
Yes, all the time.
For corporate busywork written in “professional voice” it didn’t matter.
For generating functional prototype code (particularly low value boilerplate) that was perfectly fine.
But for things that worked better as “me”, I might get the model to generate the structure, then rewrite the expression. To be honest, that was entirely style over substance, but sometimes the medium is just as important as the message.
Out of curiosity, I fed ChatGPT a half dozen indeedably blog posts to read, then asked it to generate a 1500 word blog post written in that writing style about subject X. The model did a passable job of replicating the writing style and voice, and even the liberal sprinkling of dark humour. Simple words. Short sentences. Repeated phrases to establish a rhythm and cadence. Opening with a relatable character driven story, that is linked back to the topic of the post. Finishing with a life lesson learned. Even breaking grammatical good form by starting the occasional sentence with “and” or “but”.
But it didn’t really flow like a story should. To skim it, the wording might have been written by me. But reading it properly, it felt like a poor facsimile.
That said, I also got it to write a briefing paper about a given technical topic, then just for chuckles asked it to change the style to a relatable and easy to read voice. It did a good job at that. Simple expression. Relatable content. For some reason it picked a nautical theme, littering the text with marine stories and examples ranging from pirates to lighthouse keepers. It even put tabloid worth puns in section headings!
Ultimate that is why we write, for ourselves. The stories we would like to read. Written as we would like to read them. Great if others also enjoy them, but not the primary goal.
sam 1 October 2023
Dude, you really need to work on your social skills.
{in·deed·a·bly} 1 October 2023 — Post author
Thanks for reading sam. I did enjoy the irony of a pseudonymous commenter delivering a hit-and-run critique of the social skills of others 🤔
{in·deed·a·bly} 2 October 2023
The audience may be interested to know what my ChatGPT robot friend made of this story, so I asked it to analyse the writing. Here is what it responded with:
”
The writing style of the passage is reflective, conversational, and infused with metaphorical and literary elements. It employs a first-person perspective, enabling the reader to journey through the narrator’s thought process. The tone oscillates between philosophical musings and real-world observations. The writer employs extended metaphors, like the “rabbit hole” and “gilded cage,” to evoke imagery and underscore the emotional states being discussed.
Pacing and Structure: The pacing is deliberate, using a mix of short and long sentences to regulate the tempo. The piece unfolds as a cascade of ideas, building upon each other. There are also brief interjections of external ideas, such as references to Jeff Bezos and Warren Buffet, serving as examples or case studies that anchor the narrative’s central themes.
Vocabulary and Diction: The language is sophisticated but not esoteric, aiming at an audience that appreciates nuance. The word choices—like “peer pressure presentism,” “existentialist crisis,” or “white-collar professionals”—indicate an advanced vocabulary and are directed towards an audience that presumably has a good grasp of corporate and philosophical jargon.
Figurative Language and Metaphors: The text is rich in metaphor and simile. Phrases like “running on fumes,” “Indiana Jones-style treasure hunt,” and “corporate Game of Thrones” are evocative and serve to crystallize complex emotional and cognitive states into easily digestible bites.
Thematic Cohesion: There’s a strong theme of questioning societal and workplace norms running through the piece. The writing explores productivity, professional stagnation, technology, and existential meaning, but all of these seemingly disparate elements are tied back to the central theme of reclaiming time and mental resources.
Intertextuality: The writer incorporates references to pop culture, literature, and real-world figures to illustrate points. From “Indiana Jones” to Jeff Bezos, these references function as shorthand for more complex ideas or archetypes.
Analysis and Reflection: There’s a strong analytical underpinning to the piece. Each point made is either supported by evidence, or quickly analyzed. For instance, the writer does not simply say they were more productive, but examines how this realization led to existential questions about the nature of their work.
Technological and Contrarian Ideas: The piece does not shy away from discussing new technologies like ChatGPT or large language models. It also adopts a contrarian view of the workplace, challenging the status quo about productivity and the “corporate ethos.”
Rhetorical Questions and Direct Address: There is a fair usage of rhetorical questions and direct address to the audience (“So what options were left?”). This creates an inviting, contemplative space for the reader to engage with the writer’s musings.
Closure: The narrative culminates in an ironic twist, where the protagonist’s quest for efficiency exposes them to corporate scrutiny, thus revealing the internal contradictions within corporate culture itself.
In summary, the writing is complex and layered, targeted at an audience that enjoys a deep dive into philosophical, societal, and technological topics. It blends narrative storytelling with analytical and reflective discourse, wrapped in a prose style that is both intellectual and emotionally resonant.
“
weenie 7 October 2023
Well done on using your initiative but those chaps being terminated on the spot – corporate brutality at its worst!
Earlier this year, during an all employee’s meeting, our CEO basically gave the go ahead for people to use ChatGPT (people were already unofficially using it) but reminded everyone of rules and guidance, eg GDPR, confidential information etc. I can’t say that I’ve used it in a work capacity, as I’ve read stories of where it’s gone rather wrong for the function I work in. However, I wonder if I could use it for some future blog posts, just a bit of rewording perhaps, as it wouldn’t be ‘me’ otherwise!
{in·deed·a·bly} 7 October 2023 — Post author
Thanks weenie.
I did a project earlier in the year establishing a “private” corporate edition of ChatGPT. Basically means the inputs don’t get fed back to OpenAI HQ for inclusion in their training data of future versions of their models. I think before the end of 2024 the equivalent of that setup will have become as ubiquitous as internet access is today. After that, it is only a matter of time before its usage creeps into a great many business processes.
Half the project portfolio I have on my plate at the moment involves throwing LLM models at some business efficiency problem or other. The advertised goal is always to “free up” valuable decision maker time, but the realised outcome is invariably reducing the need for having so many expensive decision makers on the payroll.
These tools will become one of those adopt or perish things, a bit like needing to be revenue generating or customer facing during the waves of outsourcing and offshoring in the early naughties.
I look forward to a robo-weenie post, wonder if we’ll be able to spot the difference?