Prediction season is here. Most predictions are self-serving hype or obvious extrapolations.

Here are my actual predictions for 2026, with confidence levels.

AI Predictions

AI Coding Assistants Become Default (Confidence: 95%)

By end of 2026, not using AI coding assistance will be like not using IDE features. Just weird.

Every major IDE will have it built in. Developers who refuse to use it will be slower than peers.

This isn’t controversial. It’s already happening.

AI Agents Disappoint (Again) (Confidence: 80%)

Autonomous AI agents that “do everything” will continue to underdeliver. Too unreliable for real-world complexity.

Specific, narrow agents for defined tasks will work. General-purpose agents won’t.

The hype cycle will crash. The useful applications will persist.

AI Regulation Arrives (Confidence: 70%)

EU will implement meaningful AI regulation. US will follow with lighter touch. Australia will copy something.

Impact on startups: Compliance costs for AI features. Documentation requirements. Some use cases become legally complicated.

AI Commoditization Accelerates (Confidence: 85%)

Frontier model capabilities continue converging. The gap between GPT-4, Claude, and open source shrinks.

What matters: Application, fine-tuning, integration. Not which model you use.

Model provider moats weaken. Application-layer moats strengthen.

Startup/VC Predictions

Funding Stays Tight (Confidence: 75%)

No return to 2021 conditions. Valuations stay “reasonable.” Due diligence stays rigorous.

Exception: AI-native companies with real traction. Still overfunded.

Good for founders long-term. Discipline produces better companies.

The “AI Wrapper” Shakeout (Confidence: 90%)

Hundreds of companies are thin wrappers around LLM APIs. Most will fail as features get commoditized.

If your moat is “we call GPT-4 with a good prompt,” you don’t have a moat.

Survivors: Those who built real product value beyond the AI layer.

Remote Work Normalization (Confidence: 85%)

The return-to-office push fades. Hybrid becomes standard. Fully remote remains viable for tech.

Companies mandating five-day office work struggle to hire. Market forces win.

Tech Stack Predictions

TypeScript Everywhere (Confidence: 90%)

TypeScript becomes even more dominant. New JavaScript projects without TypeScript become rare.

Full-stack TypeScript (Next.js, tRPC, Prisma) is the default for new startups.

Postgres Remains King (Confidence: 95%)

Despite new databases launching constantly, Postgres stays the default choice.

Extensions like pgvector add capabilities. Managed Postgres (Neon, Supabase) reduce ops burden.

Pick Postgres. You won’t regret it.

Serverless Grows Up (Confidence: 80%)

Edge computing and serverless mature. Vercel, Cloudflare Workers, Deno Deploy become standard for new projects.

“Where do you run this?” stops being a complicated question for most apps.

Kubernetes Becomes Invisible (Confidence: 70%)

Startups stop thinking about Kubernetes. Platform-as-a-Service abstractions handle it.

You’ll use Kubernetes without knowing it. Nobody will brag about their K8s setup.

Business Predictions

SaaS Pricing Evolves (Confidence: 75%)

Pure per-seat pricing struggles. Usage-based and hybrid models become standard.

AI features force the change. Variable costs require variable pricing.

Privacy as Feature (Confidence: 65%)

Some consumers start paying premium for privacy. Not mainstream, but a viable niche.

B2B: Data privacy requirements in procurement become standard.

Bundle Wars (Confidence: 80%)

Microsoft and Google push bundling hard. Stand-alone SaaS tools feel pressure.

Response: More startups focus on doing one thing extremely well. Specialist beats generalist for serious users.

What I’m Probably Wrong About

Predictions are hard. Here’s what might not happen:

• AI regulation could stall in political gridlock
• A breakthrough could make agents actually work
• Economic conditions could dramatically shift VC appetite
• Some new platform shift could change everything

The meta-prediction: Something significant will happen that nobody predicted. It always does.

How to Use Predictions

Don’t bet your company on predictions. Use them as inputs to scenario planning.

“If AI agents work, how does that affect us?” “If funding stays tight, what’s our plan?” “If regulation arrives, are we compliant?”

Think through scenarios. Make flexible plans. Adapt when reality reveals itself.

The future is less predictable than prediction articles suggest. Including this one.