What to expect in 2026
My highly opinionated predictions for 2026
It’s January 2nd, 2026; and here is the obligatory prediction piece to add to the hundreds that you are probably seeing on your feed right now. These are less crystal ball predictions and more observations of what I’m seeing in the industry for the last couple years.
I believe 2026 will be the year where a lot of trends that have been building for the last couple years will start to crystallize, crash and/or pivot.
A lot of these trends anda predictions are linked to AI, simply because is the most pervasive trend in the industry right now.
The Productivity Numbers Don’t Add Up
2025 was the year of AI first companies, CEO from all kind of verticals gave up to the FOMO and declared in 2025 their company is going to be an AI company. Executive teams then scrambled to figure out what the hell that meant for their teams and how to “make it happen”.
Every week I saw another headline about AI coding tools delivering 30%, 50%, 100% productivity gains; and every week I talked to CTOs who are quietly confused because they’re not seeing it. Not at that scale.
METR ran a randomized controlled trial — experienced open-source developers, real tasks, proper methodology. The developers using AI tools took 19% longer to complete their work. But they believed they were 20% faster.
A 40-point perception gap. It matches what I’ve been hearing anecdotally for a year.
Google’s 2024 DORA report found something similar. Every 25% increase in AI tool adoption correlated with a 1.5% drop in delivery speed and a 7.2% drop in system stability. The Faros AI Productivity Paradox report — analyzing telemetry from 10,000+ developers across 1,255 teams — showed that teams with high AI adoption complete 21% more tasks and merge 98% more pull requests, but PR review time increases 91%. The bottleneck moved downstream.
I use AI tools. They’re genuinely helpful for certain things — boilerplate, exploring unfamiliar APIs, rubber-ducking problems; creating product prototypes, etc. But “helpful for certain things” is a long way from “10x productivity revolution.”
My prediction: It’s going to be a rough year for a lot of companies that made headcount decisions based on AI productivity assumptions will start noticing their velocity metrics don’t match their shipped features. Some will quietly start hiring again. Most won’t talk about it publicly, some of the smaller ones will go out of business.
The Junior Developer Problem Is Already Baked In
2025 was a weird year overall for the hiring market, in the software industry we saw a lot of companies posting open positions but not hiring. Developers encountered one of the most difficult hiring markets in recent history.
Continuing with the AI first trend, the most impacted type of roles were the junior developer roles.
Entry-level hiring at the 15 biggest tech firms dropped 25% from 2023 to 2024. Entry-level job postings overall fell 60% between 2022 and 2024. CS graduate unemployment hit 6.1% — one of the highest rates across all majors, higher than art history or philosophy.
The bootcamp market tells the same story. Code Fellows, Epicodus, Momentum Learning — shut down. 2U bought Trilogy for $750 million and then just closed it. Didn’t even try to sell.
This isn’t a problem we can fix later. The people who didn’t get hired as juniors in 2024-2025 won’t magically become available as seniors in 2029. The pipeline doesn’t work that way.
I’ve hired bootcamp grads, career switchers, people from unexpected backgrounds — accountants, line cooks, history majors, etc. Some of the best engineers I’ve worked with. But they needed mentorship, runway, patient investment in their growth. The industry is systematically eliminating those opportunities.
Stack Overflow put it simply: “If you don’t hire junior developers, you’ll someday never have senior developers.”
My prediction: The senior developer shortage of 2028-2031 is already locked in. Companies cutting junior roles now are creating an artificial scarcity they’ll pay premium rates to fill later.
This shortage of junior developers is going to be a problem that will compound in the next couple years; 2026 will continue to become increasingly hostile for junior developers.
The Technical Debt Accumulation Is Unprecedented
Technical debt is a staple of software development, it’s a fact of life. Some companies are better at managing it than others, but they all have it. Agentic coding is making it easier to generate code fast, but it’s also making technical debt to skyrocket to unprecedented levels.
GitClear analyzed 211 million lines of code and found an 8x increase in duplicated code blocks since AI tools became widespread. Copy-pasted code now exceeds refactored code for the first time since they started measuring. Code churn — quick rewrites of recently added code — is climbing toward 7%.
Kin Lane, 35 years doing API work: “I don’t think I have ever seen so much technical debt being created in such a short period of time.”
AI makes it easy to generate code fast. That creates pressure to ship fast. Review gets rushed. Problems don’t surface until they’re expensive to fix. The 91% longer PR review times from the Faros data — that’s teams trying to catch the problems. The teams not taking that time are accumulating debt they don’t know about yet.
My prediction: The Saas world might be facing a tech debt crisis in 2026; not only are companies racking up technical debt at an unprecedented rate, but in a lot of cases they are doing it blindly, without a plan to fix it or even manage it.
This will have a few consequences:
We will see a raise in cybersecurity breaches, as more and more companies are running AI-generated code in production without enough human review.
We will continue to see an increase in vibecoded companies all through 2026.
In 2026 and 2027 we will start seeing increasing demand for senior developers not build product features, but fix AI-generated code. “AI cleanup specialists” will become a real thing.
The Bubble Dynamics Are Real
As software developers, we are trained to spot patterns. One of the patterns that got a lot of discussion and speculation in 2025 is the bubble dynamics in the AI space.
Right now we are seeing a lot of circular investment patterns that don’t make sense from a traditional business perspective:
Nvidia invests in OpenAI.
OpenAI uses the money to buy Nvidia chips.
The revenue Nvidia reports includes money they essentially gave to their own customers.
Rinse and repeat.
Michael Burry called it out: true end demand is “ridiculously small” compared to the capital flowing around.
OpenAI lost $5 billion on $3.7 billion in revenue last year. Deutsche Bank projects $143 billion in cumulative losses through 2029. Sam Altman admitted the $200/month ChatGPT Pro tier loses money because power users’ queries cost more than $200 to serve.
MIT’s study on enterprise AI: 95% of GenAI pilots fail to deliver measurable P&L impact. Pilot abandonment jumped from 17% in 2024 to 42% in 2025.
My prediction: I don’t think 2026 will be the year the bubble bursts, but it will be the year we start seeing cracks, we are already seeing some of this happen to OpenAI in 2025 as they have lost their once healthy lead over their competition, and continues burning cash like there is no tomorrow.
The most likely scenario is that 2026 will see a significant amount of consolidation in the AI space as weaker players get acquired or go out of business. Conversely, investors will become more cautious about pouring money into AI startups without clear paths to profitability.
Vibecoding and the Agent Plateau
“Vibecoding” — describing what you want and letting AI generate it — works surprisingly well for prototypes, internal tools, one-off scripts.
It breaks down for anything you need to maintain, extend, or debug six months later. The code works, but nobody understands why it works. When it breaks, you’re often better off regenerating than fixing. Fine for a one-off script. Disaster for production systems.
Cursor’s own CEO warned that vibecoding “could create weak foundations in software systems.”
On agents: MIT reports 95% of enterprise AI pilots fail to scale beyond proof-of-concept. Salesforce found agents achieve only 55% success rate on professional CRM tasks. Probability of an agent successfully completing 6 sequential tasks in 10 consecutive runs: 25%.
Enterprise reliability requirements are usually around 99.9%. Nowhere close.
My prediction: Agent adoption stalls in 2026 as companies internalize the failure rates. The category splits: narrow, well-defined tasks work fine; broad autonomous agents remain vaporware. “Vibecoded” applications built in 2024-2025 start collapsing under production loads and having major security incidents.
My Predictions at a Glance
To recap here are my five predictions:
The Productivity Reckoning: Companies that made headcount decisions based on AI productivity assumptions are in for a rough year. The gap between perceived and actual productivity gains — a 40-point difference in METR’s study — will start showing up in velocity metrics that don’t match shipped features. Some companies will quietly start hiring again; the smaller ones that can’t course-correct may not survive.
The Junior Pipeline Collapse: The senior developer shortage of 2028-2031 is already locked in. Entry-level hiring dropped 60% between 2022-2024, bootcamps are shutting down, and 2026 will continue to be increasingly hostile for junior developers. The people who didn’t get hired in 2024-2025 won’t magically become available as seniors later — the pipeline doesn’t work that way.
The Technical Debt Crisis: SaaS companies are facing unprecedented technical debt accumulation, much of it invisible. We’ll see a rise in cybersecurity breaches from AI-generated code running without adequate review, a continued increase in vibecoded companies, and by 2026-2027, growing demand for senior developers to fix AI-generated code rather than build features. “AI cleanup specialist” will become a real role.
The Bubble Cracks: 2026 won’t be when the AI bubble bursts, but it will be the year we start seeing cracks. OpenAI has already lost its healthy lead over competitors while continuing to burn cash. Expect significant consolidation as weaker players get acquired or shut down, and investors become more cautious about AI startups without clear paths to profitability.
The Agent Plateau: Agent adoption stalls as companies internalize the failure rates — 95% of enterprise pilots fail to scale, agents hit only 55% success on professional tasks. The category splits: narrow, well-defined tasks work fine; broad autonomous agents remain vaporware. Vibecoded applications built in 2024-2025 start collapsing under production loads and experiencing major security incidents.
Conclusion
Now that I’m at the end of this post, I realize just how much AI dominates the conversation, and how much it’s potentially reshaping our industry and the broader economy.
What is shocking is how far apart perception and data have drifted. The perception says: AI will progressively replace developers, productivity gains are substantial, scaling continues indefinitely, enterprise adoption is inevitable. The data keeps pointing to a different reality: productivity gains are smaller than perceived and may be negative in some contexts, enterprise pilots fail at 95% rates, and we’re destroying the talent pipeline that will need to maintain whatever we build now.
I could be wrong about all of this. January prediction pieces usually are.
But if I’m right about even half of it, the surprise of 2026 won’t be “AI is more powerful than expected.” It’ll be quieter questions: Why didn’t enterprise adoption materialize? Where did the experienced developers go? Why can’t we maintain this codebase?
In any case, here is to a 2026 that will be full of suprises and excitment, if you agree or disagree on anything I’ve said I would love to hear from you in the comments.
Cheers!