tl;dr — The year of the voice interface and job losses.
Claude Code Changed Everything
I’d been hearing about everyone moving over to Claude Code since around November, and I finally started trying it on the 18th of December. Wow. Genuine game changer. It’s not that it’s a single step change in capability, more that we’ve reached a tipping point — Claude Opus 4.5 can handle longer running tasks over much larger codebases, and it all just works.
Where I work, we don’t have a monorepo. So I just run Claude Code at the directory above and it handles everything like a champ. Multiple repos, different languages, the lot. Cursor are in trouble. I’ve signed up to the Claude Max Plan and it is 100% worth it.
The Claude Code Moment for Other White Collar Roles
Here’s what I think is the bigger story. What Claude Code does for software engineering is about to happen for other white collar roles. Sales, legal, marketing — they’re all having their “Claude Code moment” in 2026. That point where AI tooling goes from interesting novelty to genuinely indispensable. The people who lean in early have a massive advantage.
Mass Layoffs Are Coming (But Nobody Will Say Why)
Tech companies are making mass layoffs in 2026 because of AI. They won’t be explicit about the reason. Some dress it up as “over-hiring from the post-COVID correction” or restructuring for efficiency. But the reality is simpler than that. It’s not that the roles disappear entirely, it’s that AI tools combined with good people mean everyone can easily do twice as much work. So companies get away with half as many people.
Protective Measures Will Start Appearing
Some US states and EU countries introduce protective measures around human employment. We’ve already seen the early murmurs of this, but 2026 is when it gets real. Governments will be forced to confront the displacement.
The First $1B Solo Acquisition
We see our first $1B acquisition of a startup with just one, maybe two, employees. The leverage AI gives a single talented person is absurd. Someone builds something genuinely valuable with a tiny team and a big exit, and it makes headlines everywhere.
SaaS Copycats and Lifestyle Businesses
SaaS clones and copycats will be rife. When it costs almost nothing to build software, the barrier to launching a competitor drops to near zero. But the flip side is interesting too — there’s a boom in “lifestyle businesses” launched off niche software. People solving their own problems, shipping fast, and making a comfortable living from it. Not everything needs to be venture-scale.
The SaaS Moat Question (I Still Think I’m Right)
I said this last year, and I still think I’m right. Entrenched SaaS companies with deep integrations, network effects, and years of accumulated data hold their ground. But new SaaS startups? The window closes. The only ones likely to thrive operate in heavily regulated industries — finance, healthcare, legal — where compliance complexity creates a moat that AI alone can’t cross. Everyone else is building on borrowed time.
AI Enterprise Security and Data Access Is Huge
This one doesn’t get enough attention. Every enterprise wants to plug AI into their systems, but the moment you connect a model to internal data, you open up a massive surface area around security, permissions, and compliance. Who has access to what? How do you audit what the AI sees? How do you stop it leaking sensitive data in a response? Companies like Anthropic, OpenAI, and the cloud providers are racing to solve this, and there’s a huge opportunity for startups that focus specifically on AI security, data governance, and access control layers. This is going to be one of the fastest growing areas in enterprise tech this year.
Voice Becomes the Primary Interface
Finally, 2026 is the year we move from text to voice as the primary way we interact with AI. We’ve been typing prompts like it’s 2005 and we’re searching Google. The models are good enough now, the latency is low enough, and the voice quality is natural enough that talking to AI feels like the obvious way to do it. Text won’t go away, but voice becomes the default for most people. That changes everything about how we think about AI products and interfaces.