About Tomorrow’s Blueprint

Increasing productivity has been the driving force of all human progress for millenia, but the world’s labor productivity has been stagnant for decades. Even computing + internet + smartphones has led to only average growth, which is now again essentially flat. I believe intelligent applications will restore productivity growth to high levels, leading once again to significant increase in human flourishing. It’s my mission to bring that world about, and it’s why I named this website Tomorrow’s Blueprint.

What are intelligent apps?

My definition of an intelligent app is “software that is better than you are.” This is not software to digitize your paperwork and track your budget, but to automate your paperwork or give you good suggestions to improve your budget. This is not programming a robot to work an assembly line, but programming a robot to fold and put away laundry.

It’s popular to call this “artificial intelligence” today, but any term will do. The key point is that you will not know and could not predict what the software will do exactly.

How I hope to get there

I could never build every intelligent app necessary to signficantly improve humanity’s productivity. My goal instead is to create pieces of the infrastructure and software to enable many others to create the apps they need.

Hardware and infrastructure + algorithms + developers, we need all three. The recent runaway success of GPUs and Transformer models seem like they may be the breakthroughs we need for a world of intelligent apps. Although software developers are also improving (getting more productive with AI assistance), we haven’t had the breakthrough in this category yet. I expect we need most people to become “developers.” Custom GPTs was an excellent attempt in many ways, but isn’t there quite yet. Claude artifacts was another, but lack the usage.

I am at Microsoft now, because I believe Microsoft is best positioned with Copilot in Windows, Web, and Office to turn users into developers. The hardware and algorithms are there; let’s get everyone building the apps that are intelligent for them (even if they don’t realize they are developing).

About me

You may think that attempting to have a material impact on the world’s productivity is ludicrous. But my hero is The Man of La Mancha. Don Quixote never turned down a challenge because it was too hard, and neither have I. For the last 15 years, including 12 years as a Product Manager at Microsoft, I’ve been on a mission to drastically improve the world’s productivity through AI. I’ve been building the systems and infrastructure pieces necessary, and the time is now ripe to build the product: Copilot for Microsoft 365. This is my story.

Abram using a computer in college
Getting started in Computer Science, ~2008

I started right out of college deep in the belly of servers, performance, and hard disks. My first role as a product manager was establishing database replication for Exchange Online, preparing for the launch of Office 365. This was a cloud scale problem, through and through. With my engineering team, we scaled up Exchange Online by three orders of magnitude. Soon I was the DRI for tens of thousands of servers and exa-scale data.

With Exchange Online operating at cloud scale, I tackled business continuity and disaster recovery. With my help defining system features and establishing operations processes, Exchange Online moved from a two-nines product to a three-and-a-half nines product. This is the kind of infrastructure I believe is a pre-requisite for intelligent apps.

In 2015, Office 365 realized the long-foreseen potential of Artificial Intelligence was about to become a reality. As the sole product manager, I started Office 365’s Machine Learning platform. I started with 13 talented engineers, growing the charter and the team to what is now more than 200 SDE and 50 PM, with many more data scientists.

Ever since, I have been picking up the most impactful spaces within Office 365’s AI platform, staying specialized in creating new products and bringing them to viability. For example, I launched Microsoft Graph Data Connect, where I needed to establish an entirely new business model and sales channel for Microsoft 365. This was a tall order, but I knew that data would be key to the future’s AI, and I wanted to make sure every business could use theirs.

Unlike the Man of La Mancha, the windmill I’m tilting at is about to fall. Join me in drafting the blueprint of tomorrow.