The AI Speed Report: 20 Tools, Six Months, One Clear Winner

Most AI tool reviews tell you how good a product is today. That's a problem, because in this market, today is basically obsolete by the time you finish the free trial.

So we did something different. From January through June, our team logged every meaningful update across 20 mainstream AI products — release cadence, changelog depth, whether new features actually worked on day one, and whether the tool got faster, cheaper, or smarter over time. Roughly 340 changelog entries later, we've got a pretty clear picture of who's building and who's coasting.

How We Scored Shipping Speed

Three metrics, weighted equally:

• Release frequency — how often the product shipped user-visible changes (not just backend "stability improvements")
• Feature substance — whether updates solved real problems or just added confetti
• Follow-through — did announced features actually launch, and did they work?

A tool that ships weekly but breaks things constantly scores lower than one shipping monthly with rock-solid updates. Speed means nothing without reliability.

The Speed Demons

Perplexity took the crown, plain and simple. It shipped visible improvements in five of the six months we tracked, including a major upgrade to its research mode that genuinely changed how we use it. When we had a question about a feature in March, by April it had already been rebuilt.

Cursor came in a close second. The AI coding space is brutally competitive, and Cursor clearly knows it — updates landed every two to three weeks, each one addressing real user complaints from forums and GitHub. That feedback loop is what fast shipping should look like.

ElevenLabs deserves a nod too. Voice quality improved measurably between February and May, and the team rolled out new languages and cloning controls without ever breaking existing workflows.

Surprisingly Solid Middle Performers

Notion AI doesn't ship flashy updates, but it ships consistently. Small, useful refinements landed nearly monthly — better search, smarter writing suggestions, fewer hallucinated answers. It's the Toyota Camry of AI features: never exciting, never disappointing.

Canva's Magic Studio followed a similar pattern. New templates and AI design tools appeared on a steady schedule, clearly tied to its event-heavy user base (think Teacher Appreciation Week and Q4 holiday campaigns).

The Coasters

Now the uncomfortable part. Two well-known products shipped essentially nothing user-facing during our entire six-month window beyond bug fixes and pricing changes.

We won't name the worst offender, but here's the tell: if a tool's last "What's New" blog post is from over eight months ago, and its changelog reads like a list of excuses, you're paying for a product on autopilot. In one case, a competitor's free tier now outperforms the paid plan we tested — which is a tough look.

Jasper landed in a gray zone. It shipped regularly, but many updates felt reactive — chasing whatever OpenAI released that month rather than building anything distinctive. If you're paying a premium over raw model access, you should be getting more than a wrapper that updates when the wrapper's API does.

What This Means for Your Wallet

Before you renew any AI subscription, spend ten minutes on the company's changelog or release notes page. Ask yourself:

1. When did the last real feature ship?
2. Are updates addressing user complaints or just adding features nobody asked for?
3. Is the roadmap public, or is the company silent between billing cycles?

In a market where new challengers appear weekly, stagnation isn't neutral — it's decline. The tools winning right now aren't necessarily the ones with the best launch. They're the ones that treat shipping as a habit, not a milestone.

Our advice: subscribe to the fast movers, trial the middle tier, and cancel anything that's been quiet for a quarter. Your invoice will thank you by summer.