Building Agents that will work in production in the Enterprise - a great framework and example case from Ankur Bhatt on how to think about building agentic capabilities into enterprise workflows. Another great read on this topic - Weighing value vs. risk is the key and inside an enterprise the risk side is non trivial. This is the key point to keep in mind “”most agent failures will be boundary failures, not model failures....The agent story will continue to be told as “intelligence.” The winners will build constraint, revocation, and recoverability without fanfare. Control will become the product.”

Median retirement saving in American is $955 - This when Baby Boomers are reaching retirement age, retirements are lasting longer, jobs for older Americans are shrinking, Medicaid spending (the primary funder of assisted living for poor) is being cut and Social Security is already strained.

Social safety nets (SNAP and Medicaid) are getting harder to get for those who are unemployed. As we make our way through another layoff season this will become more painful to watch.

A theoretical story of AGI progress written April 2025 that is tracking pretty well as of Q1 2026.  An interesting progression forecast story that is worth pondering (with a couple options on the ending).  TL;DR It doesn’t end well based on the premise that the AI begins to recursively take over its own training and outpaces the ability of humans to keep up.

A solid AI mental model discussion on the what and how AI works that gives you some context to better separate the paranoia from the probable.  (and a bit of a counter point to the Something Big is happening article below).

A beautiful read on Tool Shaped Objects (with a solid counter point to the adoption curve of AI automation) but most importantly it highlights why it’s urgent to have clarity on what we use tools for. I wrote up a POV here and I liked Tim O’Reilly’s take on Abundance here

Anthropic Agent Autonomy Research - This is clearly an “early days” research but it suggests what we already know - that software engineers are doing the most with agents and that the range of agent autonomy is wide but also generally small, in other words this is all early days.  The suggestion to not make universal friction in process was interesting to me, I can understand that at the macro it is likely to add unnecessary friction.  I am very curious why there was no discussion about the need to highlight the risk/reward to the user.  As more “non technical” people begin to use agents there is a real need to protect people from the unintended consequences of automation.  Anyone who has accidentally deleted something important, or created a run away recursive process, knows how easy it is to scale a mistake.  I care less about adding restrictive assumptions to any process and much more about elevating some understanding on the reality that the agent is going to do what you ask well beyond the boundary of what you might have considered in the request.

Where are the new moats - Jake Saper has a great thesis on this - I am generally in agreement with this list and Amy Wilson are discussing the topics of moats and value creation in Season 2. I think there is a key point in standing behind the results - this is over and above the pricing topic and toward the trust and context topics.

Productivity gains with AI - the real value creation is the second order effects as the coordination costs shift (h/t Reshuffle) - “AI does not just make existing organizations more efficient. It makes the organizations themselves less necessary.” What does this mean for value creation indicators.  We’ve talked on the show about how our heuristics of leadership strength are mis-aligned to this future when they use size of org as a proxy for skill,  Information asymmetries are emerging in new places and that means that value capture is as well. The new 10x employee has much more leverage than in the past.

Private Equity trapped capital - a great scan on the current state of Private Equity. As SaaS businesses become strained in the public markets this is a topic to watch as the exit pipeline becomes challenged (if you want an explainer on that part this is a great read)

America isn’t ready for what AI will do to jobs - The Atlantic makes clear the limitations of economists to predict future as the entire discipline is about looking at past data and patterns. This is important not just for economic understanding but for any disruptive/chasm crossing moment. There are important moments where past performance do not predict future success - we are in one.

Nate B. Jones es explains why your domain expertise is not sufficient to protect you from the job collapse from AI.  - Meg POV: what we must build now is the ability to respond to massive change. To do that effectively, we need to learn to consider multiple possible futures, and build the technical foundation to act on whichever one turns out to be true. Waiting to be certain is the worst choice, and the one that comes with the highest opportunity cost.

Something big is happening - Meg’s POV: "It's not an emergency until it happens to you" - Meg's Grandmother.  The volume is getting very high in tech and it has an important lesson for all industries. The future will be different from the past, and the future is coming at us faster every single month (not year). It's no longer "if" but "when". The advice is the same no matter the context. There has never been a more important time to have curiosity and a beginners mindset. Getting hands on with the latest tech and learning how to adapt with the changes is critical, building financial resilience is also urgent. If you are looking for ideas I'd personally merge those two topics and get to work.

A counter point from Gary Marcus on what the above gets wrong.

Do NOT (I repeat do not) piss off or even offend our new agent overlords. A cautionary tale. h/t Benn Stancil

A counter point to high agency - “There’s an intense pressure to be an unusual person who will be unique and get the funding.” - Meg POV: I feel that there is an equivalent horseshoe theory for the agentic v. mimetic. While I am of the belief that high agency matters, at the extremes it seems massively at risk of the realization of an extremely sad end state.  Some time back Benn Stancil and I had a comment thread where we considered the reality of today where the goals of money and fame seem to be replacing the past tech innovation cycles that were all about “making the world a better place”

Bot takes down AWS by deleting code (see my Reading and Pondering point 7 above)