From PCPRO November 2025
Here’s our glossary of the future to help you decode everything from “algospeak” to “vibe coding” – so when you hear these terms in meetings, you’ll be able to figure out what the C-suite actually wants.
Agentic AI or AI agents These versions of AI can actually take action, rather than merely generate reams of text or quirky images. Thankfully, because AI remains stuck in our computers that action is largely limited to navigating the web and filling out digital forms. If these agents work, AI might become useful for taking over our digital drudgery. At least until they take our jobs.
AI washing Marketing your product as having AI when it doesn’t really, or only in a limited way. You know how every company claims to be AI native or AI first these days? That. See also: GPT-washing, jumping on the bandwagon, AI hype, every day since ChatGPT arrived.
Algospeak According to Adam Aleksic’s new book on the topic – aptly named Algospeak – this refers to words created to evade algorithmic censorship on social networks, or any term otherwise shaped by algorithms. For example, Chinese-owned TikTok didn’t like users chatting about death or other related topics; instead, they switched to the term “unalive” to dodge algorithmic controls. See also: leetspeak, euphemism treadmill.
Artificial general intelligence Commonly referred to by its acronym AGI, this is a term with an always shifting meaning. DeepMind co-founder Demis Hassabis says it means AI with “human-level cognition”; OpenAI CEO Sam Altman says AGI is “the equivalent of a median human that you could hire as a coworker”, though in his company’s contract with Microsoft it’s defined as when AI can make $100 billion in profit. See also: shifting goalposts.
Bare-minimum Mondays The opposite of starting as you mean to continue, this TikTok work trend is all about easing into the work week by spending the first few hours of your Monday organising yourself (read: doing laundry) and strategising (read: taking a nap) rather than feeling flustered (read: doing work). See also: Sunday scaries, lazy-girl jobs, quiet quitting.
Brain rot This internet slang is applied to low-quality digital content, often found on social media, which is useless yet weirdly addictive, as well as the cognitive decline caused by such material. Used in a sentence: “I got tired of doomscrolling, so tried to relax with brain rotting instead – that was a skibidi move.” See also: doomscrolling, bedrotting, lay down and rot (LDAR)
Broligarchy A portmanteau of “bros” and “oligarchy”, this trend relates to tech billionaires wielding their power over society and politics, and found its perfect example/nadir in Elon Musk’s foray into the United States government with DOGE.
See also: techno-feudalism – how the world has always worked, but with this time around terrible fashion sense.
Clanker A derogatory term for AI or robots, derived from the Star Wars cinematic universe. Used in a sentence: “I called customer service, but there were no humans to talk to – just clankers.” See also: wireback, cogsucker, tin can.
Cloud repatriation Remember back when servers were physical devices in the building, rather than something you accessed via AWS? Well, the good old days are back! Companies angry at cloud costs and their unpredictability are re-investing in their own on-premise systems again.
Context window How much an AI model can think about or remember, sort of like its memory. A wider or larger context window means more information can be considered, leading – hopefully – to improved accuracy. Used in a sentence: “If you want a larger context window, you’ll have to sign up for a paid subscription.”
Decel Short for decelerationist, a derogatory term for those who think AI development should be artificially slowed down to help avoid, or help us humans prepare for, all the dangers that AI-developing companies keep warning us about, like an existential threat to humans.
Delve A signifier for AI content, “delve” is one of a handful of words that generative AI systems use more than we humans, so its appearance suggests AI writing. Of course, some people do actually use that word, and now that AI uses it frequently, humans are probably going to start using it more, sparking a spiral of “delve” overuse that means we can no longer use it to spot machine-generated words. Another example is the punctuation device known as the “em dash” – beloved of professional writers, including this one, which is probably why AI uses it so much.
De-monetisation When influencers are shadow-banned by the algorithms, reducing how much their content is shown to would-be viewers, hurting their chances to push unnecessary products to their followers. It’s being digitally cancelled, but just on socials.
DePIN Short for decentralised physical infrastructure networks, this is blockchain in the real world, a way of applying that as-yet not very useful innovation to manage and control real-world assets such as sensors or edge computing.
It’s the internet of things meets crypto. Yes, we know.
Digital sovereignty The US is no longer trustworthy – if it ever was – so it’s time to invest in our own cloud, AI or other technical services that are in our physical control; essentially, this means local data centres.
Alternatively, a way to get Microsoft, AWS, Google et al to invest in local infrastructure. See also: cloud sovereignty.
Techno-feudalism: how the world has always worked, but with this time around terrible fashion sense
Echoborg When a human’s speech is entirely scripted or determined by an AI, this originally referred to a research technique for social psychologists but now also explains why the person that applied for that entry-level graduate role took ten seconds to reply after every question in the Zoom interview and said “delve” three times.
Frugal tech Sustainable, affordable innovation that’s designed to benefit people, frugal tech is seen as an antidote to the dominance of Big Tech, broligarchs and Silicon Valley. Used in a sentence: “I read a great feature about frugal tech in The Guardian, but still bought the new iPhone.” See also:calm technology, dumb tech.
Generalisability An as-yet-unrealised goal for automation, in particular driverless cars, generalisability means a technology that can be used anywhere without much – or indeed any – localisation effort. For example, a self-driving car trained heavily in San Francisco could hit the streets in LA without years of work and billions of dollars in additional investment. Or a humanoid robot could see the dishwasher needed emptying and figure it out without smashing the crockery. See also: unicorns.
Generative engine optimisation Like search engine optimisation (SEO), but GEO is targeting the attention of AI-powered search rather than humans. Bots talking to bots all the way down. See also: dead-internet theory.
GitOps DevOps, but on GitHub. Similar insider-indicating jargon built on the DevOps style includes ChatOps, LLMOps, ModelOps, MLOps, AIOps and really anything else with too much hype. All seem to be leading to NoOps, which removes humans from the loop using automation, though there’s pushback via HumanOps. (That was initially written as a joke but appears to be a real thing.)
Glazing When an AI sucks up to you by saying your writing or ideas are really awesome, in a bid to keep you using the machine. “That’s a great question, Hal.” See also: sycophancy, AI psychosis.
Guardrails Rules built into an AI system in an attempt to prevent it from spitting out scary responses, such as violent, biased, overly sexual or racist images and text. After all, being nasty is a job for us humans.
Hallucination A dramatic way to describe generative AI errors that suggests inaccuracies are mistakes rather than a key indicator of how these systems actually work, which is by making things up. Used in a sentence: “We’ve made significant advances in reducing hallucinations, improving instruction following, and minimising sycophancy,” said OpenAI announcing GPT-5.
Native Not an acknowledgement of indigenous rights, but a way to jump on the technology bandwagon in order to sound ahead of the curve; useful for AI-washing. When the backlash against future tech begins, we’ll surely be talking about “human-native” systems.
Used in a sentence: “Our AI-native security app used to be called cloud-native, but marketing told us this would sell better.”
Meeting recovery time A measure of how long it takes to get back to actual work after a meeting. This is why you get so little done on days with too many Zoom calls or all-hands. See also: Zoom fatigue, meeting-free mornings/days.
p(doom) The probability of doom or hope, or how likely you think AI will prove an existential risk and kill us all. See also:AI Doomerism, decelerationists.
Prompt engineering Learning how to talk to the machines so they understand what you actually want, despite the fact these are language models designed to work with natural speech. Secondary definition: your new job. See also: prompt tuning, tweaking until you get it right.
RTO An acronym for “return to office” that is usually accompanied by a “mandate”, an RTO is a demand from management to get out of your pyjamas and commute to headquarters rather than work from home as we have all successfully done since lockdown. The opposite of hybrid working, RTOs are big in banking circles – the money needs to be babysat in person, it would seem – and increasingly so in technology. See also: presenteeism, micromanagement, nostalgia for a long lost but not missed style of working .
Scope 3 emissions Emissions from a company are broken down into three types: Scope 1, or direct emissions; Scope 2, indirect emissions, such as the electricity used; and Scope 3, which are supply-chain emissions. The latter sound like they’re someone else’s fault, but when it comes to Big Tech this is where big rises in emissions are found: Google’s Scope 3 includes the production of technical infrastructure such as hardware for AI and has increased by 22% year on year. See also: greenwashing.
Shadow A way to describe a technology or tool that’s been snuck into a company under IT’s nose without consideration of the potential risks. That includes shadow AI, and the original, shadow IT. See also: letting ChatGPT do your work for you.
Shift left Not a political reference – sorry, Jeremy Corbyn – but the trend of moving security considerations or other features earlier into the development lifecycle of software. See also: Secure by design.
Slop The garbage content spat out by AI that’s now clogging the internet, social media and our work lives. And, in turn, worsening AI itself. Often comes in the form of unfunny memes – weirdly often including babies for some reason – but can be applied to almost anything churned out by generation bots. See also: slopification, enshittification
Slop: The garbage content spat out by AI that’s now clogging the internet and social media
Synthetic data AI models need to be fed data for training, but there isn’t always enough of the right type, so AI can be used to generate synthetic data to show that system in training. That makes sense with, say, driverless cars, as accidents or other extreme, rare incidents can be manufactured to show the AI model so it knows what to do if and when it happens in real life. But it could also mean AI-generated material is being used to train an AI to make AI-generated material, and it all gets a bit too meta for us. See also: model collapse, digital inbreeding
Synthetic workforce When all employees are clankers. See also: ghost knowledge worker, virtual human.
Trendbait Aleksic defines this annoying trend as saying something in a social post in the hopes of sparking responses to go viral. See also: ragebait, clickbait, the entirety of social media, tabloid headlines.
Vibe coding Coined by Andrej Karpathy in early 2025, vibe coding is the use of AI to create an app or other software, writing the code itself based on user prompts alone. Tell an AI what you want, it writes the code, even if you don’t know how – especially if you don’t know how.
No comments:
Post a Comment