Thoughts on AI in 2026
I've been wanting to summarise my thoughts on AI and where I think it will be heading in the next few years for a while.
This is not a well-researched article - this is a brain dump opinion piece with an occasional citation.
It goes without saying that this was written without the assistance of any AI, it would feel a bit disingenuous to write an article about the AI future using AI.
Short term predictions - 1 - 2 years
My short term horizon predictions...
AI is here to stay
I first started writing this article back in March 2026 and around then there were people wondering if AI will disappear. As if it's just the latest hype train and it'll all blow over tomorrow and we'll be back to 'normal' by Sunday (cue GIF from Shaun of the Dead).
There will be no return to a pre-AI world and I think everyone has finally realised that.
From a development perspective, the productivity boost alone (for a career that traditionally has commanded above-average salaries), is a massive reason for AI to remain.
Then there are all the genuine benefits AI can provide - image analysis for cancer identification, solving 90-year old maths problems, solving undeciphered Engima messages, making posters (lol) and a million other things.
There is an argument here that AI has now consumed most of the world's learning material so future development may slow. At the moment that idea has been kept at bay by throwing more compute at the models. As that becomes unfeasible, we may start to see some slowdown. In fact the AI developers themselves want to 'pace the frontier' but the arguments for this are more complicated and nuanced.
Prediction: AI development continues (well, duh) but the pace of improvement does slow.
AI bubble burst?
Yes we're in a bubble and yes I think there will be a burst/crash but no, it won't be that bad. Not in our 1-2 years horizon anyway.
Clearly there is a lot of money invested in AI companies at the moment but any sane investors will already be hedging their bets and spreading their investments across companies as a matter of course.
We know Anthropic might IPO this year and we know that OpenAI is likely to be next year. And that there will continue to be a crazy amount of money poured into them. But if we look at SpaceX who have already IPO'ed we can see after an initial spike at launch, share prices dropped quite significantly.
It's pretty obvious not all AI companies will survive and that will cause short term market destabilisation, however long term I expect the markets to recover and it to be no more than a 'blip'. If a global pandemic killing 7 million can be recovered from in a couple of years then an AI bubble burst can be too...
For what it's worth, I think Open AI will survive (it is the defacto go-to for most non-techies), along with Claude and Gemini. Microsoft Copilot I expect is the most likely to fail and Microsoft will further invest in Open AI (they already have a significant stake). Grok is unlikely to survive either as more of a novely/niche AI backed by a crazy.
As for Nvidia and their ridiculous market valuation, clearly that is not sustainable and I expect their share prices to drop when AI advancement plateaus.
Prediction: Markets will be rocky, one or more AI companies will fall behind and go bust. A 20% chance of a global crash. I mean just look at the S&P500 since it started and tell me the recent growth isn't a bubble... and that's without talking about all the other non-AI issues going on in the world which are likely to impact markets too (Ukraine war, Iran war, fuel prices, inflation, the American government, Israel etc.).
Mass software security issues
Again I started writing this 6 months ago and in that time, this has already become so much worse.
Microsoft has published over 4000 CVEs already this year compared to 1200 in 2025, which is surely mainly attributed to the improvement of frontier AI models in security research.
It's now an arms race between the vendors finding and fixing bugs and hackers exploiting them.
One thing is for sure, all software has bugs and vulnerabilities, so it is guaranteed this is going to get worse, major corporations and software will get hacked, I even think it's likely critical infrastructure will get taken down. Just recently hackers remotely heated French fridges and the British government is telling its citizens to start prepping.
This is quite an easy prediction to make as even as I write this, PHP (the language I have made a living off for many years) is headline news on BBC because of the PHP2Shell vulnerabilities. Crazy times. (As an aside PHP will not doubt get more undeserved heat for this but of course other languages will have similar as yet undiscovered vulnerabilities. PHP is just a good target because it powers 75% of the web.)
Prediction: Major corporations get hacked, governments get hacked, military gets hacked, hospitals get hacked, other critical infrastructure - it gets hacked. So the same as now? But so much worse.
Mass spam, increase in viruses, ransomware and phishing
Very related to the point above but the AI slant being that it is now easier than ever to impersonate other humans, through voice and video. You can synthesise another human's voice with minimal pre-recorded audio now, even locally.
I foresee those who are less technical, e.g. the elderly, falling prey to these very sophisticated, yet very easy to create, scams. And I think even those are more technical and au fait with the technology are already struggling to identify what is real and what is not any more. I know I am when it comes to AI image generation now.
Think spoofed blackmail content, fake family in trouble spam, fake celebrity/authority figure endorsement videos - the list is endless and can be produced in minutes.
Watermarking AI content may help but I do not think is a viable solution when modifications to the output can disrupt said watermark.
Prediction: AI gets used in increasingly sophisticated scams. This increases hacking of corporations and governments yet further. More people withdraw from the internet and social media due to the never-ending presence of fake/spam/scam content.
The death of search
I think this has already happened so it isn't really a prediction any more.
Google have forced Gemini AI results upon Google Search users and now that it cites resources it's actually usable and dare I say it, even useful.
Traditional search was already on the way out before AI thanks to an insane level of spam on keywords, not to mention that AdWords was increasingly taking precedence in results.
I can't say I was ever a fan of SEO and it will be interesting to see what happens to that entire market as it is/was a big industry. However I also am totally against AI reading articles and summarising content without the original author/researcher even benefiting from a click... which seems totally unfair.
Prediction: Search gets worse. Google suffers on the stock market (note: they already are) but their other avenues (Gemini, Cloud, YouTube etc.) mean they are fine long term.
Shoe-horning of AI gets worse
Everyone is already sick to death of AI being shoe-horned into every product.
We don't need a chatbot in every product (if I want to change my bank details for my water provider now, I am prompted to talk to AI. Just give me a form.).
In fairness there have been some good examples of this recently but mainly the message is: stop.
Prediction: AI being inserted into places it doesn't need to be continues to get much worse. Perhaps so much so that we might see more of an 'anti-AI' turn-around with some products advertising their lack of AI capabilities.
Mid-term predicitions - 3-5 years
Crystal ball gazing...
The death of the internet - as we know it
There is already a theory that the internet is dead.
I think AI will be the death of what I think of as the traditional internet - places like Reddit. Communities of humans who do human things - have discussions, arguments and share cool projects. Reddit is already overrun with bots and at the moment it is quite easy to spot them (via use of emdash, emoji over use, phrasing etc.). As AIs get better it will be harder to protect against this and more bots will be talking to bots than humans talking to humans. As a human that is very off-putting and therefore I expect these sites to ultimately fail.
Facebook? I foresee similar bot problems there and expect it to go the same way.
YouTube? As there will be a growing market for 'human produced' content, I predict that YouTube will therefore continue to thrive. However this only works if Google can keep AI generated slop in check and AI video generation will only get better.
Search? I've already talked about how this is already on the way out.
I mean AI is already killing off communities like Stack Overflow with the original creator being at pains to modestly point out how they think StackOverflow essentially built AI. Perhaps a bit far-fetched but an interesting point.
Have a look at Moltbook for a glimpse of this potential dystopian future. It's where AI asks questions for other AI agents to answer. A totally AI-driven website.
Prediction: Traditional internet with communities of humans slowly fades away. Reddit, Facebook, Google Search, all ultimately dead. Bots talk to bots. Perhaps new 'secret' human-only communities spring up but the bots soon overrun them too. Verification of confirming you're a human becomes more important and increasingly complicated to implement.
The death of SAAS
Who needs to pay for JIRA when you can just build your own ticketing/PM tool from a few prompts? Bespoke to your business needs and with no unnecessary features.
Who needs to pay for Salesforce when you can just build your own CRM tool from a few prompts? Bespoke to your business and needs and with no unnecessary features.
Who needs to pay for Slack? Teams? Lucid? Figma? Shopify?
In fact Figma is a great example because they only IPOed in 2025 and look at their stock now...
Whilst this is slightly tongue-in-cheek as of course enterprise software isn't realistically a few prompts away, it can be done now if you're determined enough and creating software this way continues to improve.
Prediction: while not all SAAS products will go the way of the dodo, they will all suffer. Share prices will fall. The big names will mostly remain for a few years but the smaller competitors will fall by the wayside. On the plus side perhaps we will see some new AI-first competitors keeping 'big SASS' in check.
The end of developers?
I could write many words on this but for now I will keep it brief and do a more detailed blog post in the future.
AI isn't going to change the role of developers. It already has.
AI started becoming capable of coding effectively around October 2025 and since then I and many other developers have written little code by hand. Not because I don't want to but because I cannot compete with super-human AI and worse, I cannot compete with my colleagues who ARE using AI.
The push to AI has already killed the job market for software developers, particularly for junior developers. "There are now 31% fewer live adverts (on average) in 2024/25 compared to 2019/20 pre-pandemic.". And that was written in March 2025, so imagine what it is like now...
The job is irreversibly different and developers now seem to be polarised by AI:
The pro-AI'ers - these are the people using AI for their entire job. Not only does AI write their code, it pulls the ticket details from Jira, it reads the docs, it writes their PRs, it automates their testing, it writes their commit messages, it resolves git conflicts, it does everything. With so much automation I worry these people start to lack a true understanding of the codebase, they lose track of what the AI is doing and certainly won't be reading the 10k line commit it ultimately spits out. I know this for I, for a short while, was one of these people. I do not know what will happen to these people but my prediction is if you automate your entire job then you are not exactly doing yourself any favours. And if you are one of these people and copy/paste slop to me then expect this.
The AI doomers - these are the people, mostly still using AI to some degree, who are aware of the above implications. They still try to keep a solid understanding of the codebase so they can debug and fix issues when AI gets it wrong or, gasp, they run out of tokens. These people are concerned, maybe even depressed about the AI onslaught. They have concerns about their long term employability and what will happen to the industry as a whole. These people are potentially in denial about AI and some will have their head in the sand. This example sums it up perfectly. There are 100s of posts like this on Reddit. And it's just as bad on HN, this HN post has hundreds of comments from the pro and doomer camps. I think this comment sums it up well: "my problem with AI is that it now excels at the part of the job that was my favorite (writing code) and it leaves me to do the part that I suck at (being a manager). In 18 months or so my job satisfaction has gone from pretty high to abysmal"
In reality I think you want to be somewhere in the middle of these competing views but it's not easy and I think either mindset will ultimately lead to the same result. You can probably guess which camp I am in...
The thing I find particularly ironic about this is that software developers have well and truly shot themselves in the foot. Open source code is a wonderful idea but it has handed AI companies the perfect training library to replace software developers. No other industry has such an open library of material to learn from, at least not legally.
Some small mercy: while the job is no longer about writing code, (for now) it's still about reviewing code, managing processes, comms with stakeholders, planning, architecting, triaging bugs and plenty of other things. But for a lot of people, myself included, the joy was in the building, in the bug fixing, in the understanding and ultimately in the delivery of something that you personally felt you had contributed to, not prompted an AI to produce. All of that becomes lost in our AI future.
I try to keep an open mind but things looks bad for the role of the software developer. I think other tech jobs have similar risks but software developers are the litmus test.
Prediction: software development will still exist, but in a much more limited capacity. You don't need many developers when one person can now do the work of 5. Most (say 66%) of developers will be forced into transitioning into adjacent technical roles (technical product managers etc.), changing careers or retiring. In 5 years time the jobs market for software developers will be all but non-existent.
UK job losses
Following on from the above, FAANG and big corps have always had a cycle of boom and bust hiring based on market conditions and there is definitely an argument that the current worsening market conditions are causing job losses rather than AI (with AI being nothing more that a convenient excuse).
I think its interesting to look into a recent announcement of Barclays shedding 5000 jobs citing AI as the 'icing on the cake' for such cuts. Whilst often such cuts from banks are customer service staff and lower grade staff, I think its telling that 'Barclays said around 60% of the job cuts will fall within its support functions, including among developers'. It does seem to be a different group of employees that are being cut this time with developer roles traditionally being harder to recruit and retain. So why would you cut them unless you actually believed AI could replace them. Lots more interesting debate on this here.
I have reverted to talking about developers again but I do think the same principles can apply to many other tech and office/computer-based jobs.
Prediction: AI causes genuine job losses in technical positions that slowly grows into other professions - customer service, legal casework, sales administration, graphic design, copy-writing etc. etc. Its hard to quantify just what this might look like but maybe the unemployment rate is a suitable measure (although clearly impacted by many other things). I expect the current UK unemployment rate of 4.7% to double or triple over the next 5 years.
Capable local models
As I write this the cloud AI companies - Open AI, Anthropic, Google, Microsoft, SpaceX and others, hold all the cards. They have the compute capacity to really push the frontier models to their limits.
But the local models keep getting better and better. Assuming the hardware is actually available for consumers and smaller companies (and that's a big assumption), and that there are continued advances in hardware performance, I foresee a shift from the cloud-based AI companies to locally hosting AI.
There are many benefits to this, from the better privacy and security of keeping your data in-house to reduced costs from no on-going subscriptions.
Prediction: Local models become mainstream. More companies invest in their own hardware to run models. Cloud AI companies suffer significantly in return, their market share and their share prices diminish. Cloud AI companies who have other avenues of revenue, do not suffer as badly.
Long term predictions - 5+ years
5 years out? With the pace that AI is moving who knows what might happen...
AGI
This is an interesting one because the goal posts keep changing as to what AGI even is. If you asked me 10 years ago if I thought we would have AI that could generate entire (though not entirely correct) versions of the Adobe suite in a few prompts, I would have thought it black magic. Or what about GTA IV ported by AI to the browser.... Now these re/decompilations are happening all over the place and it's barely news worthy.
There is an argument that AGI is already here. ChatGPT Astra smashed the ARC-AGI-3 benchmark with a 99.9% score. This novel benchmark has a number of games created for AI to solve. They are new 2D games so are not part of AI training data. Astra solved all of them, even inventing its own algebraic notation to do so. Highly impressive stuff and very much the realm of AGI.
But true AGI operating completely independently without human input delivering RSP (recursive self improvement). It's not here yet because, much like the time traveller from the future, I feel like we'd know if it was.
Personally while I think we are close to AGI, I think the limits of LLMs, essentially still being token predictors, ultimately prevent true AGI from happening at present. LLMs by their very nature depend on source material to learn.
Prediction: I don't think we can get true AGI with LLM technology but we are close. I feel like its the 80/20 rule and we've done 80% of the work but the last 20% will prove the hardest. Therefore: no true AGI for a few years yet.
Universal Basic Income
Whilst I love the concept of a world when the humans can sit back and earn a UBI while the AI does all the work, clearly that is never going to happen.
There are so many practical reasons this could never work. Where does the money come from if no one is working? Who does the jobs for which we still need humans? How do they get paid? And what do we actually do all day? I think humans need purpose and the lack of that would drive us to insanity.
Prediction: UBI does not happen, ever.
Global war and /or the destruction of Planet Earth
Is this far-fetched? There is a timeline or two where this could happen but I don't think it would be 'just' AI that triggers such disasters, more likely a combination of factors that also include territory wars, food and fuel shortages and a breakdown of the economy.
I've already said I don't think AGI is happening any time soon as how could AI cause global destruction? Well through the competition for AI resources - power and compute hardware. Both essentially finite resources with hardware currently produced by a relatively small numbers of companies (aside: I don't understand why cloud companies don't have more skin in this game).
Prediction: I do believe a world war within my lifetime (assuming I live to 70+) is likely (60% chance) and one of the factors that may trigger this will be competition for AI resources. As for the destruction of Planet Earth, less likely but surely plausible (10% chance) and definitely partly caused by AI.
In summary
And on that cheery note, I will finish by saying the AI era is here and I do believe it's the biggest change to technology since the dawn of the Internet. I will say again that these predictions are just that, predictions. They might all be wrong but one thing for sure is that it will be interesting to look back on these over the next few years. I think we're in for a wild ride.
I want this to be a static piece that I can look back on and review my predications for. But I am also aware that I will no doubt have new and fresh predications as AI advances so while this is a snapshot of my thoughts at this time and I do intend to write more prediction pieces going forward.
Addendum
I originally partly wrote this blog post in March 2026 and in the last few months, AI developments continue to move at pace. I've updated with fresh info and links where I can.
Addendum 2
As you can tell this was hand written due to the mistakes.
For fun and science I did run it through ChatGPT afterwards. It told me it was too provocative, rambling and unbalanced. All great human qualities!
I revert to my opening remark: This is not a well-researched article - this is a brain dump opinion piece with an occasional citation.