Chamath Palihapitiya Channel

The post-launch phase of a software platform actually accounts for 50 to 80% of its overall lifetime cost. Interestingly, the traditional service model is structured to conclude precisely when this highly expensive period gets underway.

As soon as a project is released, typical system integrators will reassign their delivery teams to entirely new clients. This practice ensures that the original creators behind your core design are completely out of the picture long before ongoing support is required and the most costly years commence. You have to wonder who will be there to assist you three years down the road when compliance regulations shift or when the foundational workflow of your system becomes obsolete.

Breaking away from this standard cycle, 8090 maintains full accountability for your production code well beyond the initial go-live date. We make sure your software remains consistently updated and perfectly adapted to your evolving business landscape.

Discover more by visiting https://bit.ly/4zhwMBA

We are looking at an incredibly positive step forward for the industry. With the financial costs of different models gradually aligning, the natural progression will involve a shift toward independent third party control planes and harnesses. Some notable examples leading this charge include 8090 Software Factory, Palantir, and Nous' Hermes.

Looking ahead over the next 36 months, it is highly likely that every enterprise will demand access to a third party harness paired with the flexibility to utilize any model quasi-interchangeably. This shift will occur as organizations become fully aware of the vulnerabilities surrounding data and IP leakage. In the end, companies need to maintain absolute sovereignty over their encoded judgement, which makes this adaptable approach essential.

On an annual basis, Gallup evaluates the distribution of employees under individual managers across the American workforce. Over the decade they have been monitoring this metric, the ratio recently experienced its most rapid shift, climbing from 11 to 12 direct reports in merely one year. The catalyst behind this trend is a familiar talking point during corporate restructuring discussions. Organizations are consistently eliminating middle management roles and allocating the remaining personnel among a smaller pool of existing leaders.

According to Gallup's research, employee engagement actually reaches its optimal level when a manager oversees a team of eight or nine people. At this size, a supervisor retains the ability to monitor ongoing projects accurately and offer constructive, genuine feedback. However, once the number of direct reports grows past fifteen, engagement drops drastically.

Today, the crucial issue is no longer just the volume of individuals assigned to a specific management title on an organizational chart. Instead, the focus must shift to understanding the proportion of an employee's week that is spent on tasks requiring complex human judgment, rather than routine duties that simply keep the business functioning.

The enterprises currently gaining a competitive edge have managed to completely eliminate repetitive tasks from the schedules of their human workforce. By doing so, they ensure that the professionals remaining on the team are dedicating their time to customer relationships, nuanced conversations that require a human touch, and the strategic innovation that ultimately drives market success.

Ways we can help your organization

Approaching the Artificial Intelligence Singularity

The prevailing theory regarding this phenomenon unfolds through a distinct sequence of events. First, humanity successfully develops an Artificial General Intelligence. Second, this newly created system achieves mastery in the field of artificial intelligence research. Third, the system leverages its expertise to engineer an even more capable version of itself. Fourth, that upgraded iteration goes on to build a remarkably superior successor. Fifth, this feedback loop continues to accelerate at an unprecedented rate.

When reviewing the latest achievements and technical functionalities emerging from multiple research facilities over the past few weeks, it becomes quite evident that we are already securely positioned within this continuous cycle.

We can expect the upcoming 18 months to be absolutely extraordinary. From this point forward, the mechanism of recursive self improvement is poised to drive up technological capabilities at a breathtaking speed. As a direct result of this rapid evolution, the marginal costs associated with operating all of these models will eventually drop to roughly ~$0.

On a positive note, we can take comfort in knowing that our entire token expenditure is being directed toward a truly worthwhile cause.

Elon was absolutely correct in his early vision for an accessible American AI. Releasing Grok as an open source project could represent a strategic checkmate for him. The reasoning here is quite compelling. By making Grok open source, the profit margins are naturally forced out of the foundational model layer. Instead, they are redistributed downward into infrastructure components like chips and CSPs, as well as upward into the application sector. Considering he already controls a massive data center build, possesses next gen inference AI, and owns Cursor, which operates as the most utilized application, this move would arguably yield a higher margin for him. Following this logic, the concept of eventually relocating these data centers into space becomes an even more practical strategy.

I am getting ready to step in as a cohost for @SquawkCNBC right now. You can catch the program during the 7-9am ET time slot. I am absolutely thrilled to get things underway!

If you have been keeping an eye on the ongoing discussion regarding American closed source systems versus Chinese open weight models, we have some exciting preliminary data to share with you.

We recently ran an initial pilot focused on upgrading an application from PHP to Next.js. The results showed that combining 8090’s Software Factory with Opus 4.8 was 1.5x faster and simultaneously 1.4x cheaper than relying on Opus 4.8 by itself.

We also tested our Software Factory alongside the more affordable GLM 5.2 model. This combination yielded an impressive 16.4x reduction in expenses, although it did operate 3x slower than using Opus 4.8 alone.

At this stage, these outcomes are purely directional, with a sample size of n=1 per arm. For our next steps, we intend to execute these trials again under proper controls. We will be utilizing 10 to 15 buyer-relevant legacy modernization tasks that are currently being designed alongside our Sales and GTM teams.

Expect more updates from us soon. In the meantime, this data prompts a rather important question regarding financial strategy. Given the massive cost advantages of open weight options, why would any American public company continue to burn through shareholder capital on expensive closed source models?

To explore this further, our upcoming tests will focus on evaluating American open source models, specifically looking at Nvidia.

David Hammons currently stands as the most important artist in the world. By pushing boundaries, his work manages to simultaneously explore historical pasts and predict future events, a difficult balance that all top creators strive to master. At 82 years old, he is still going strong and remains a highly active force in the creative community.

Anyone who appreciates art and happens to be in NYC between May 1 and June 13 should make a point to visit White Cube. The gallery is hosting a phenomenal exhibition that looks back on his incredible career. I am especially thrilled about this event because I was recently in a position to loan the gallery a piece of his art that he eventually permitted me to purchase.

The story behind acquiring that work goes back to 2014 when I spent an unforgettable morning with David inside his Brooklyn studio. It was during this visit that he shared a profound secret with me. After I requested to take a photograph together, he declined. He then instructed me to get down on the ground next to him. Looking directly into my eyes, he explained that he refuses picture requests from everyone, advising me instead to store the memory using my heart and not my camera.

That piece of advice has stayed with me ever since. By putting the device away, I am still able to clearly recall the atmosphere of his studio, the immense depth of his creativity, and the gracious way he treated me alongside everyone else. His method absolutely worked. Looking back, his words seemed somewhat blithe at the time, but they carry heavy weight today when genuine presence has become so rare. We currently live in an era dominated by taking selfies and staging artificial shots for social media, making his timeless lesson more relevant than ever.

We are seeing outstanding earnings reports from the Mags right now. Curiously, though, the equal weight index is substantially outperforming the weighted index today, a fascinating development that certainly gives us something to ponder.

I am pleased to share that my 2025 Annual Letter has officially been published and is ready for you to read at https://t.co/vjvpDz6exF

Inside this edition, I explore a wide variety of subjects. The discussion begins with an overview of returns before moving into deeper conversations about artificial intelligence. Specifically, I examine the theory that if AI ultimately destroys terminal value, the future essentially holds zero worth until that moment actually arrives. I also explore what the global consequences will be for everyone else if the US and China maintain exclusive ownership over frontier AI.

The letter further highlights how the most definite bottleneck facing artificial intelligence is entirely physical. Alongside this, I look at how breaking up Big Tech through trust-busting might be completely incompatible with the goal of winning the AI race. Other sections cover the growing trend of utilizing doomerism as a pitch to raise funds, as well as an analysis of how financial markets are shifting their pricing structures away from traditional P/E models and moving instead toward current free cash flow.

There are many more topics covered throughout the document, and I hope you find these insights valuable.

The strategic landscape surrounding artificial intelligence is experiencing a fundamental shift. While owning a top tier foundational model remains crucial, the true bottleneck is increasingly becoming physical real estate that already possesses zoning approvals and adequate power supplies. Whenever you can combine that kind of property with immediate access to silicon, you have secured an unbeatable advantage.

Right now, holding these specific resources provides extraordinary bargaining power. That leverage is only going to grow stronger as local communities continue to reject new data center proposals. Elon recently illustrated this exact point through Cursor.

Looking ahead, one can only imagine the scale of the agreements that OpenAI and Anthropic will need to negotiate over the coming years. In fact, the recent partnership between Amazon and Anthropic was merely a small preview of what is to come. For astute investors sitting across the negotiating table with the right physical assets in hand, the upcoming opportunities are exceptionally promising.

The rise of artificial intelligence is fundamentally transforming how organizations operate, effectively boiling businesses down to three essential components.

First, there is Expert Knowledge, which is the specialized, industry level understanding of exactly what needs to be accomplished. Second is Tribal Knowledge. This refers to the informal, rarely documented nuances of a business that explain why some companies achieve much better results than their competitors doing the exact same work. The third piece of the puzzle is Hardware and Software, serving as the connective tissue that enables people to fuse this expert and tribal knowledge together to get things done.

When a company manages to thoroughly document its expert and tribal insights, it opens the door to completely redesigning its daily operations. Rather than taking jobs away from people, this process actually fuels faster growth and greater job creation. As the core tasks become perfectly clear, the organization requires an ever growing number of individuals to step in and apply critical human judgment.

This exact philosophy is the driving force behind our creation of Software Factory, a solution that many large and established enterprises rely on today.

At present, the greatest danger facing modern social networking applications stems from the powerful combination of an exceptionally advanced video model, TTS technology, and automated research capabilities.

An exceptionally advanced image model represents the single greatest danger to the competitive advantage currently maintained by Instagram.

To be entirely frank, the current re-rating of Nvidia defies logical explanation. Additionally, I find the shifting FCF multiple assigned to Apple, Microsoft, Meta, and Alphabet to be equally baffling. Regardless of my confusion over these specific adjustments, it is quite apparent that the underlying pricing models for public markets are going through a swift evolution at this time.

Please feel free to check out the latest video release by visiting this link: https://youtu.be/J2qKVjRR12w

We are currently experiencing an astonishing rate of progress and development. Because the business landscape is moving forward so rapidly, organizations that fail to adapt and evolve stand out much more clearly than they have in the past.