Another year begins, another orbit complete. The nacelles are stable, the consoles are humming, and the view-screen is full of familiar patterns repeating themselves — just at a slightly different scale. If 2025 was about correcting course, 2026 looks like a year of consolidation, containment, and a renewed appreciation for systems we can actually understand and control. This isn’t a post about moonshots. It’s about gravity; that’s my lame way of saying don’t expect any bombshell surprises here.
1. Linux Desktop: Slow, Steady, and Finally Boring
Projects like Omarchy and COSMIC aren’t trying to reinvent the workstation. They’re doing the unglamorous work of making Linux feel cohesive, performant, and predictable for people who actually ship software. That matters more than market share charts ever will. These projects have taken a great leap in making Linux on the desktop (especially for pros) occupy more mindshare and indeed market share. Developers, operators, and technical leaders are increasingly choosing Linux desktops not as a political statement, but as a professional tool:
-
More similar to their actual deployment environment (ie Linux)
-
Tighter control over updates
-
Better alignment with containerized and server-first workflows
-
Less friction between “how I build” and “where it runs”
This won’t show up as a hockey stick in consumer metrics, but it will quietly reshape engineering teams. Of course, as it goes more mainstream there will be some pushback against the popularizing projects, but I don’t see this having any real negative effect outside of social media / Reddit nonsense.
2. AI Is Consolidating — and It’s Going to Get More Expensive
The AI gold rush phase is coming to an end. What comes next looks a lot like every other infrastructure cycle we’ve lived through. Fewer vendors. Higher prices. Tighter contracts. We’re already seeing it:
-
Models consolidating behind a handful of providers
-
Usage caps replacing “unlimited” marketing copy
-
Per-token pricing quietly creeping upward
-
Features moving behind enterprise-only tiers
Training frontier models is capital intensive; that’s a fancy way to say it costs too much money. Serving them at scale is even more so. The era of subsidized inference is ending, and 2026 will be the year many teams feel that shift directly in their budgets. For leadership, this creates a familiar tension: AI is now considered mandatory, but it’s no longer cheap or frictionless. That pressure will force a harder look at where AI actually delivers value versus where it’s just been bolted on because “we should be using AI.” A good example of this is the recent Warp pricing changes that are in effect for most users significant increases. Which leads us to the inevitable next step.
3. The Rise of Self-Hosted LLMs (Behind the Firewall)
As AI consolidates and prices rise, something very predictable happens: folks will start bringing it in house. Even before the price increases truly hit in earnest, there’s already significant privacy concerns with the current AI as a service model.This process of self-hosting will for most organizations likely start with Internal tooling
-
Codebase-aware assistants
-
Documentation and search
-
Test generation and diagnostics
-
Domain-specific reasoning
- Internal organization-wide knowledge-bases
Crucially, they run where your data already lives — behind the firewall, on hardware you control, under policies you understand and can audit. For many organizations, especially in regulated or industrial environments, this isn’t a downgrade in a any way. It’s an upgrade. Latency becomes predictable. Costs become knowable. Security reviews get simpler instead of harder. In Star Trek terms, this is the difference between renting time on someone else’s starship and finally maintaining your own warp core.
LLMs are maturing, organizations are getting wise to the costs and risks associated with them and the AI vendors need to stop burning money faster than a Klingon on a barrel of blood wine. All in all, 2026 is shaping up to be a year of maturing transition and maturing practices.
