Leadership CONNECT: 27-JUNE-2025 (Y25W27)

The Value of Deep Dive

As of today, I have captured 979 thinking Sparks in my Box “AI or AGI.” Of these, I can say I looked deeply (meaning spent more than 5 minutes) at about 20%. I read and digested (meaning I spent more than one hour on each) 5%.

The Spark “State of AI in 2025” is now my most valuable read and digested item on AI in the past 30 days.

It is an excellent summary because it:

1. Balances business, social, and technical aspects.
2. Combines data and insights, many of which I agree with, and many of which were new.
3. Well-organized (PDF and Video), with direct links to sources.
4. Includes great zoom-ins if desired.

In short, if you are an AI thinking leader, I recommend reading the PDF [1] (or watching the video [2] for more details).  

Key Seven Insights by Davis TLDR

1. Generative AI has gone mainstream – 1 in 8 workers worldwide now uses AI every month, with 90% of that growth happening in just the last 6 months. AI-native applications are now well into the billions of annual run rate.
2. Scaling continues across all dimensions – All technical metrics for models continue to improve >10x year-over-year, including cost, intelligence, context windows, and more. The average duration of human task a model can reliably do is doubling every 7 months.
3. The economics of foundation models are…confusing – OpenAI & Anthropic are showing truly unprecedented growth, accelerating at $B+ of annual revenue. But, end-to-end training costs for frontier models are near $500M, and the typical model becomes obsolete within 3 weeks of launch, thanks to competition & open source convergence.
4. Just like the smartest humans, the smartest AI will “think before it speaks” – Reasoning models trained to think before responding likely represent a new scaling law — but training them requires significant advances in post-training, including reinforcement learning & reward models. Post-training may become more important than pre-training.
5. AI has now infiltrated almost all specialist professions – From engineers and accountants to designers and lawyers, AI copilots and agents are now tackling high-value tasks in virtually all knowledge worker domains.
6. Agents finally work, but we are early in understanding how to build AI products – Agents have finally hit the mainstream, but design patterns & system architectures for AI products are still extremely early.
7.  “AI-native” organizations will look very different – Flatter teams of capable generalists will become the norm as generative AI lessens the value of specialized skills. Many roles will blur – such as product, design, & engineering. 

Video Summary

1 in 8 workers now use AI monthly. AI-native apps are pulling in billions in revenue. But under the surface, the pace is even more dramatic – models are getting 10x cheaper, faster, and smarter every year… and becoming obsolete in just weeks.

Davis Treybig (Partner at Innovation Endeavors) unpacks what’s really happening across the foundation model landscape – and what’s coming next. From model performance and agent design to startup economics and emerging use cases, this is a practical guide for anyone building or investing in AI. 

Contents:

0:00 – Intro

0:58 – Overview of past 5 years

9:08 – State of foundation models today

36:00 – Use cases and applications

44:43 – Building foundation model products

1:05:34 – Market structure and dynamics

1:13:40 – What’s next?

1:21:10 – End

Super Cool Things I Liked (Yesha’s Thinking)

• Slide 7 – Scaling models leads to “emergent” behavior – These LLMs are beginning to understand and perform tasks on their own, and we do not fully grasp why and how.

• Slide 12 – LLMs quickly surpass almost all new benchmarks as they are released – AI surpasses humans — now accept it. See  [3].

• Slide 18 – Compare GPT-54 and DeepSeel-VL – Same quality for 10% of the cost.

• Slide 96 – OpenAI is becoming a consumer app company, and Anthropic an API company — I didn’t know that. I thought it was the other way around.

Note: The slide number may change as the deck is updated. Here is a copy of the PDF [4]

P.S. If you are an AI-thinking leader, spend the hour reading [1] or viewing the video. [2]

More Links

• [1] Download the full deck here: foundationmodelreport.ai
• [2] Watch the video 1 hr 20 min
• [3] “Machines Can Think Better Than Humans” So What?
• [4] Stored PDF version of the PDF