We are now living in the future that Microsoft's researchers were asked to predict. In late 2015, Microsoft assembled sixteen of its top researchers and asked them to forecast what technology would look like in 2016 and beyond, spanning AI, cryptography, data science, and human-computer interaction. What makes this link compelling in May 2026 is that we can finally do a full retrospective scorecard — not on vague, safe predictions, but on specific claims made by named researchers with reputations on the line.

This kind of artifact is rare and valuable for several reasons:

• Calibration check on expert forecasting. The tech industry is saturated with predictions, but we almost never go back and grade them. This piece gives us a clean dataset: specific claims, specific timeframes, made by domain experts at one of the world's largest research labs. How well did they do? The answer tells us something important about how much weight we should give to expert technological forecasting in general.
• The AI predictions are the most interesting to revisit. In 2015, deep learning was ascendant but the transformer architecture hadn't been published yet. GPT didn't exist. The concept of an LLM as a general-purpose reasoning tool was not on most researchers' radars. Whatever the Microsoft team predicted about AI almost certainly undershot what actually happened — and the specific ways they undershot reveal the blind spots that even top-tier researchers had about the trajectory of their own field.
• Cryptography and security predictions age differently. Unlike AI, where progress was explosive and nonlinear, advances in cryptography and formal verification tend to be more incremental and predictable. Comparing accuracy across domains within the same prediction exercise highlights which areas of computer science are fundamentally more forecastable than others.
• It's a mirror for our own predictions. Right now, every major lab is making bold claims about AGI timelines, robotics breakthroughs, and quantum computing. Reading what smart people got wrong a decade ago is a useful exercise in epistemic humility before we take today's forecasts at face value.

The piece is also a snapshot of Microsoft Research's priorities before the Satya Nadella-era OpenAI partnership reshaped the company's entire AI strategy. It captures a moment when Microsoft's AI ambitions were still largely academic rather than product-driven, making it a subtle piece of corporate history as well.