.svg)
THE BIG IDEA
This is how you save 4,500 developer years with AI
GenAI might be hitting the Trough of Disillusionment inside many companies, but one tech leader is seeing eye-popping ROI — and it might be a roadmap for others to follow.
Amazon CEO Andy Jassy is singing the praises of Amazon Q, their new generative AI assistant designed to tackle one of the most dreaded tasks in software development: updating foundational software. Earlier this week Jassy raved on LinkedIn about how Amazon Q has drastically reduced the time required for Java upgrades, cutting what used to take 50 developer-days down to just a few hours. This efficiency boost has saved Amazon the equivalent of 4,500 developer-years—yes, years—of work.
In just six months, Amazon has upgraded over 50% of its production Java systems to modernized versions at a fraction of the usual time and effort. Even more impressive, 79% of the auto-generated code reviews were shipped without any additional changes. The benefits extend beyond developer time savings, with enhanced security and reduced infrastructure costs translating to an estimated $260 million in annualized efficiency gains.
The true potential of specialized LLMs like Amazon Q lies in their ability to improve over time as they are more widely used. As these models learn and adapt, they unlock increasingly significant value for enterprises by becoming even more efficient and effective in performing complex tasks. Jassy sees this as a game changer for large enterprises, with plans to expand Amazon Q's capabilities further to help developers streamline even more tasks.
But the impact of Amazon Q goes beyond just efficiency. In a leaked chat, Matt Garman, CEO of AWS, told employees that AI's growing role could fundamentally change what it means to be a software engineer. Suggesting that as AI takes over more of the coding work, developers will need to shift their focus from writing code to understanding customer needs and designing the end product. "It just means that each of us has to get more in tune with what our customers need and what the actual end thing is that we're going to try to go build," Garman said.
Looks like we all might get the chance to do those "customer-centric" things we always say we'll do during quarterly planning before the crushing reality of each sprint comes into view.
Garman isn't the only one imagining how engineering will change in the AI age. Marco Argenti, CIO of Goldman Sachs, has also spoken out about the evolving role of engineers in the age of AI. In an article for Harvard Business Review titled, Why Engineers Should Study Philosophy, Argenti argued that technical skills alone won't be enough to set you apart in the AI era. Instead, engineers will need to excel at breaking down problems, thinking from first principles, and even "debating a stubborn AI." These are the skills, he suggests, that will define “a great engineer in the future.”
First up, though: It's time for more corporate leaders to copy Amazon's playbook and find highly-specific use cases where GenAI can eliminate rote work and actually make engineers' lives easier. Then, we can all become Socrates.
CHART OF THE WEEK
The Future of Work in the Age of AGI
Could AI outpace human intelligence by 2027? Leopold Aschenbrenner, a former member of OpenAI's Superalignment Team, thinks so, and he warns that enterprise leaders, tech innovators, and product builders need to prepare now.
In his 165-page analysis of AI's future, Aschenbrenner stresses the importance of situational awareness as we approach the era of AGI and superintelligence. He underscores the need for scaling compute resources, bolstering security measures, and developing advanced alignment techniques to navigate the profound changes expected in the coming years.
Aschenbrenner foresees a "techno-capital acceleration," with trillions of dollars being directed toward AI infrastructure—GPUs, data centers, and power supplies—to support AGI development. For tech leaders, this means the race is on to build and scale the infrastructure necessary to unlock AGI’s potential. Early investments in these areas could position companies to dominate their sectors if AI-driven products become the dominant growth engine.
However, a significant challenge lies ahead—the diminishing availability of high-quality internet data for training larger models. This suggests that companies will need to innovate in data generation and model training to maintain the momentum in AI advancements. As AGI evolves from simple chatbot-like interfaces to sophisticated, agent-like systems, the nature of work will inevitably change. Roles involving high-level cognitive tasks could be redefined or even phased out, requiring enterprises to rethink talent strategies and prioritize upskilling and reskilling.
While Aschenbrenner's analysis is eye-opening, it's important to remember that he's both an AI investor, and deeply embedded in the industry's effective altruism movement. He and many at OpenAI and beyond believe AGI is the natural evolution of today’s generative AI models. Yet, the broader consensus among experts suggests otherwise—arguing that AGI might not be the inevitable outcome.
CLIENT SPOTLIGHT
20 years in, Buzzback Innovates $100B+ Research Market
The market research industry is on track to hit $108.57 billion by 2026 — how could an industry veteran, like Buzzback, strike boldly into this quickly-expanding market?
Market research is equal parts quantitative and qualitative. Recent years have delivered innovation for quantitative research but the qualitative side has struggled to catch up. It remains “manual, fragmented, and slow,” according toBuzzback's Liz White.
That’s when White saw a massive opportunity: One streamlined platform, called Studio, that would connect market research experts with the organizations who need them and give them the software they needed to collaborate. If done well, Studio would help Buzzback expand with its Fortune 100 and Fortune 50 clients.
But to bring this bold idea to life, Buzzback needed more than a traditional dev shop. They wanted to treat the Studio build like a mini startup: super lean, using only the most critical resources required to reach the finish line. That made A.Team’s model extremely appealing.
Studio’s zero-to-one build required four A.Team members, which they onboarded in just one week: a lead engineer, a product lead, a front-end developer, and a designer. A.Team’s approach felt unique and exceptionally transparent. According to White, “Other agencies might bring their full dev team, but you don’t know who is actually working on the project. With A.Team, we were given control of choosing the exact right people for the work.”
In just one year of work, Studio launched to rave reviews. Users found the experience "seamless," and Buzzback was thrilled with the quality of A.Team, with White stating, “Here’s what I would scream from the rooftops: A.Team’s quality was leaps and bounds above some of the development firms with whom we’ve previously worked.”
“With a lot of other development firms, the onus has been on us to spec out a project.” Said White, “By contrast, everyone on our A.Team was a true collaborative partner.”
Post-launch, White considers the A.Team partnership a resounding success—so much so that Buzzback is rethinking how it will hire technical resources for future development projects.
WATERCOOLER
University Student Builds Nuclear Fusion Reactor in His Bedroom with AI
Looking for a new unexpected AI use case for Claude? How about building a nuclear fusor from scratch in your bedroom. Hudhayfa Nazoordeen, a mathematics undergrad at the University of Waterloo, did just that—with the help of Anthropic's Claude 3.5 Sonnet AI assistant. Despite having zero prior hardware experience, Nazoordeen spent a week diving into design principles and getting familiar with tools and components from McMaster-Carr. Over the next few weeks, he assembled the fusor, turning his bedroom into a DIY nuclear lab and documenting his journey on twitter.
DISCOVERY ZONE
Afraid of the dark? A startup called Reflect Orbital is now selling Sunlight.
MEME
.jpeg)
.jpg)
.png)
