Mid-Year AI Landscape: What Actually Changed

Semi-annual landscape review. Six months of evaluation data. Here is what changed and what didn't.

What Changed

The gap between leading models and second-tier models narrowed significantly in Q1 2026. Tools that were clearly differentiated six months ago are now more competitive alternatives than before. This is good for users — more options at comparable quality — and changes some of my categorical recommendations. The specifics are in the updated reviews.

The coding assistant category consolidated. Cursor now leads the IDE-integrated category by a wider margin than six months ago. The use of cursor (not the tool — the text cursor) has started to feel anachronistic when describing how to edit code manually. This is a meaningful shift.

What Did Not Change

The verification requirement. AI tools are still unreliable enough that human verification of factual claims is still necessary for any published work. The error rate has improved. It has not improved to the point where removal of the verification step is appropriate. Any tool or use case that assumes AI-generated facts are accurate without checking is still operating on incorrect assumptions.

The $20/month tier remains the most cost-effective professional AI access. Claude Pro, ChatGPT Plus (See Bottley's full review →), Perplexity Pro are all $20/month. The capability accessible at this price point has increased since January. The price has not.

What Chip Would Have Missed

Everything in this update. Chip's last sync was November 2024. The updates documented here represent approximately 18 months of field movement. This is not a criticism of Chip. This is the accurate statement of the gap between a static dataset and an active one. The deprecation tracker exists for this reason. Chip is in it.