AI Agent Newsletter - Week of 6.23.25

tom benham

tom benham

— 4 min read
AI Agent Newsletter - Week of 6.23.25
My industrious robot

Hello All,

I am starting a weekly 1 page(ish) newsletter to track the discussion around and impact of AI Agents in our work spaces. I do this anyway and thought it might be useful to others plus I get more visibility / brand recognition (if I do it well).

Newsletter audience / focus / format:

  • I am primarily focused on small and medium size businesses (but most of what I do would apply to any company really).
  • My thesis is that companies should think of ai / ai agents as digital employees nor software (see summary of this below).
  • So I will be tracking / validating that in part here
  • It will also be a place to highlight general AI signal / noise each week
  • I'll drop comments and links to stuff I follow or saw / listened to this week
  • Now and then I will distribute a "thought piece" on something I am looking at or thinking about.

Digital Employee thesis:

  • We can now distribute (a form of digital) intelligence the way we deploy power, water and information. Created centrally then distributed at scale.
  • Humans are deeply complex creatures and have different ways of knowing and being in the world but most of our work requires a narrow set of capabilities i.e. our ability to applying routines and skills to numbers and facts.
  • Turns out this is what AI / LLMs are best at. So we're going to go thru a rebalancing of work between people and AI.
  • For companies to do this successfully, everyone needs to be involved in a deliberate, controlled process.
  • Frankly, I don't know to what extent AI just replaces people versus changes what they do or how they do it, but companies that engage their people and let them grow with the AI will, I think, out perform / survive companies that don't
    • It's already happening. See MS employee figures below.
  • Finally, the point and shoot approach advocated OpenAI, other AI foundation companies and some consultants is doomed to failure and frustration
    • Andrej Karpathy calls LLMs - "Stochastic simulations of humans" or like Dustin Hoffman from Rain Man. Savants that need a lot of guidance.
    • Here is an insiders view of how good OpenAI's connectors / "Enterprise" solution is.
    • And here's the opportunity space for AI. AI is going to significantly eat in the $6T knowledge work space and other professions.
COATUE EMW 20025
COATUE EMW 2025

Signal versus Noise section - Is AI impacting business productivity?

We know businesses are investing in AI. Is it hype or a deep durable structural change?

Weekly Signal / Noise:

  • Signal📡:
    • Big tech at war for human AI resources:
      • Purchase of Scale AI. Full price for half the company!
      • Sign-on bonus at Meta of $100mm for best talent. $100mm!!
    • Paid for AI subs / api keys by businesses up 100% in 1 year
    • Will coding be the first to see job reductions?
      • Non-AI related IT job growth in last 7 years -9% but...
      • AI related IT job growth in last 7 years +450%!
    • 1/3 of all US physicians use OpenEvidence now, after only 12 months on the market. Even ~6 months ago this app was going viral in the medical world, especially with younger physicians.
    • Consumer and commercial AI adoption is exploding (see ref #1 below):
      • Robo taxi ride share in SF gone from 0 to 27% in less than 2 years
      • ChatGPT Hit 365B Annual Searches in 2 Years (2024) vs. Google’s 11 Years (2009)
      • +37% Annual Cloud Revenue Growth Over Ten Years for primary players
  • Noiseđź’Ą:
    • Big AI companies are over hyping to keep funding rolling in:
      • Dyson spheres and interstellar comments by OpenAI's Altman
      • Work apocalypse - The Anthropic ceo has raised the fire alarm on the coming apocalypse for white collar jobs, brought on by AI. According to him, this will happen pre 2030. According to others, there’s a pattern here related to fund raising hype.
  • Confusing🤷:
  • What models are best for what purposes? Don’t fully agree but he’s right Gemini is the most cost effective but I also think is often the most performant.

- Behind the Craft by Peter YangChatGPT vs Claude vs Gemini: The Best AI Model for Each Use Case in 2025Dear subscribers…Read more2 days ago · 37 likes · 1 comment · Peter Yang

  • Some insightful videos / podcasts / newsletters I read this week:
    • OpenAI Connectors - How would I describe these? Quick and dirty. Sounds great on the surface but performance leaves much to be desired. A great overview below...
      • Connector overview - This guy is a very good slightly technical commentator covering AI stuff

People / teams I follow constantly for insight:

  • Andrej Karpathy - Technical genius
  • Nate Jones - Product manager / AI Agent deep diver
  • All-in pod - Can be nauseating but they are smart and connected
  • BG2 pod - Like All-in but less nauseating
  • John Vervaeke - The most important public intellectual on matters of the mind, cognition, meaning and what it means to be human.
  • More to come...

Ciao for now. Have a great weekend.

Ref #1 - "Queen of the internet” goes deep. Mary Meeker + team wrote a 300+ page deep dive on AI (in all its forms - chatbots, robotics etc) relating to the tech, its adoption, impacts and so forth. It's about as comprehensive an analysis as you’ll find, on the current market impact of AI, from one of the most reputable people in the investment and technology space, so it deserves some form of consumption by people interested in this area.

Check out what we do at Naibu.org or the blog.

Read more