MKT Studios

Battery Spencer

Mon, 28 Sep 2026 00:00:00 GMT

One bridge, multiple angles, always beautiful. One of my first experiences viewing it was from Battery Spencer. The high vantage point, the breathtaking views of the Pacific, and the San Francisco skyline in the background are all things far better experienced than described.

No Authority, Full Context

Mon, 21 Sep 2026 00:00:00 GMT

This is the first essay in a series of AI 101 posts that I’ll be writing. In each of these, I’ll break down topics and concepts related to Artificial Intelligence into digestible chunks with a product manager’s twist.

Today, when you enter a prompt into an AI like Claude, ChatGPT, or Gemini from providers like Anthropic, OpenAI, or Google, you’ll see it think for a bit and return a response. In reality, what you’re typing into is just a thin client that doesn’t do anything significant by itself. Like most user interfaces, it’s a facade that abstracts the technical complexities underneath. The response actually comes from several complex layers beyond what you see, and how it comes about looks a lot like how a product gets built.

The harness

The next layer your prompt goes to is the Harness that lives hidden from plain sight on your device, or on the servers of your organization or provider.

It is similar to a product manager in many ways. A product manager’s responsibilities include:

Understanding explicit and implicit customer and business needs.
Producing supplementary artifacts that better clarify the problem.
Drafting the product requirements document (PRD) including the problem and the artifacts.
Communicating the PRD to a cross-functional team for development.
Ensuring that the product gets clearance for shipping.
Shipping the finished product to users.

The harness does all of these.

The primary role of the harness is to draft the Context (PM analogy: PRD).

To do this, it gathers the following:

Your newest message and conversation history. (Customer needs)
A set of high-level instructions pre-baked into the harness by the provider, called System Prompt. (Business needs)
A dictionary of tools that the harness can invoke, including ones from third parties created via the Model Context Protocol (MCP). (Supplementary artifacts)
Documents that the harness can retrieve to augment the generated response (RAG). (Supplementary artifacts)

And while gathering, the harness optimizes context for ideal relevance and length by trimming or summarizing older history and retrieving relevant chunks of information, all of which have a direct impact on quality of output and performance of the system. It also indicates which portion of the context is cacheable to the provider, further boosting performance. And when this is done, it passes the context to the Model (cross-functional team) via its API.

The harness has no real authority over the model and purely influences its output with a powerful context. The model is also stateless. It retains no part of the context and requires the entire context—including the content marked cacheable—to be passed from scratch again. Think of this as a product manager updating requirements mid-development, except that, in this case, the cross-functional team always requires a full PRD rewrite for every iteration.

Once the model processes the context, it generates a response and passes it back to the harness. The harness ensures that the entire process stays within the guardrails set by the provider (clearance). Think of guardrails as caps like content safety concerns, security permissions, maximum tool calls, time consumed, or usage credits. And finally, when the harness determines that the response is ready, it presents the response back to you via the interface where it all started (shipping product).

The spectrum of prompt complexity

Your prompts to an AI are not always simple. They could fall anywhere across a spectrum of increasing sophistication as in the following examples:

“What’s Tokyo like to visit in November?”
“What’s the current exchange rate from USD to JPY?”
“Find me a flight to Tokyo under $800 for the second week of November, check if it conflicts with anything on my calendar, and book it if it’s clear.”
“Plan my Tokyo trip: research flights, find a hotel near Shibuya under $200/night, and put together a 5-day itinerary based on my interest in food and temples — then combine it all into one plan.”
“Plan and book my entire Tokyo trip: first find flights. Only search hotels near whichever area the chosen flight actually arrives at. If total cost of flight plus hotel exceeds $1,500, drop to economy-tier hotel options and re-search. Once both are confirmed, check for calendar conflicts, and only finalize if there are none. Log every booking attempt.”

The most basic chatbot

Simple prompts like (1) are completed in a single run of the harness-model interaction. The harness passes the context to the model; the model returns a response that the harness surfaces to you. These are how Chatbot harnesses work. They can answer simple questions that just require a single harness-model roundtrip but can do nothing more.

The chatbot that does one more thing

Slightly more complex prompts like (2) require a little more than fetching a response from the model. These require invoking tools available to the harness such as the ability to search the internet, check the weather, execute some code, or some third party capability made available via the Model Context Protocol. So, in this flow, the harness reaches out to the model with the context. The model responds asking the harness to run a specific tool and to use the tool’s output in the response to you.

The need for agents

Significantly more complex prompts, such as (3) through (5) above, require much more work. These not only include tool runs (2) but also Loops, Orchestration, and Graphs.

Loops are iterative harness-model interactions (3). Orchestration involves managing multiple parallel but independent iterations (4). And graphs are multiple sequential iterations where the next step is conditional on the previous step’s outcome (5). All of these require one or more agents.

Agent is the name of a harness in a loop—drafting, sending, receiving, and revising—repeating its core process multiple times to arrive at the ideal response. Just like how product managers can’t always complete a PRD in a single run, harnesses too might require multiple runs (PRD iteration) to get to the best possible response.

In each loop, the harness updates the context, feeds the model, receives the response, and updates the context again until it thinks the response is ready for you or it hits a guardrail set by the provider. This loop is called the ReAct loop or the “Thought → Action → Observation, repeating”¹ cycle.

The agent can update the context in several ways. It can use the output that comes from executing tools. Or it can also spawn multiple sub-agents that each update the context.

Sometimes, these sub-agents work independently with the parent agent orchestrating the operation. Think of this as a product leader who manages a portfolio of unrelated products with a team of product managers.

Otherwise, these sub-agents work sequentially and based on some specific condition being met. These related sub-agents comprise a graph. Think of this as a large enterprise product that has product managers owning individual components. Product updates could require changes to multiple components. This work sometimes has to be done sequentially; for example, update the foundational platform framework before its dependent clients can implement. And the components that need rework would vary based on findings from the prior set of changes. These related components form the graph that lives under the hood of a large enterprise product.

The spectrum of agent autonomy

Consider two other prompts:

“Find flights to Tokyo under $800 for the second week of November. Don’t check my calendar or do anything else without asking me first — confirm before each step.”
“Plan and book my entire Tokyo trip — flights, hotel, and a 5-day itinerary — based on my usual budget and preferences. Handle everything end to end; I don’t need to review anything unless something goes wrong.”

Prompt (6) explicitly requires the harness to check in prior to each step. Prompt (7), however, gives the agent full autonomy. There are several prompts that would sit in between, granting varying levels of autonomy to the agent. Generally, the need for agent autonomy/human involvement is determined either by the prompter, the provider, or the model itself.

Explicit prompting

If the prompt clearly specifies check-in points, the harness follows that instruction and seeks direction for next steps. So in prompt (6) above, the agent does what it is asked to do and checks in before proceeding. This is the simple case and is similar to a product manager iterating on the PRD with further input from the user.

Harness policy

Sometimes the model’s response could require certain actions from the harness, like accessing and updating a file on the system or running some command via a tool, that the harness is restricted from carrying out. These restrictions are explicitly built in as a guardrail by the provider. In such cases, the harness checks in with the prompter, seeks appropriate permissions, and proceeds based on the consent received. Consider these similar to stakeholder reviews seeking sign-offs for key product decisions.

Model’s judgment

Even if the prompt and the harness allow for full autonomy, the model’s response could require the harness to seek explicit permissions. While explicit prompting and harness policy are deterministic—they are based on explicit conditions—the model’s judgment is not guaranteed; it is probabilistic, as with most things concerning the model.

The PM parallel, if I have to, admittedly, stretch the analogy, would be the cross-functional team seeking user feedback on a build because they are conflicted on some decision.

Every response you get from an AI goes through one or more product development cycles. The harness uses the context to influence the model without authority, and updates the context using user and stakeholder guidance amid ambiguity. This is all too similar to how a product manager uses a PRD to build a product with a cross-functional team.

Most of the time, the shipped product is only as good as the final PRD that stands behind it. Writing a complete PRD end-to-end requires heart: you need to be able to fully empathize with your customers; it requires mind: you need to think of all the use cases the product must cover and to help your cross-functional team see what you see. And most importantly, it requires a lot of time—the most valuable resource for us humans. I once wrote a PRD for software that recorded inbound inventory at a grocery store. I had to work with a global engineering team, store operations, and my end users—the store workers and managers themselves. I had to fully understand how the stores operated physically as well as how the current software worked, including its design and architecture. I just had to write the PRD once and make occasional updates based on new findings as the team went through redesign. And that entire process took me months. The harness writes its PRD, the context, from scratch for every iteration of the response, and it does that significantly faster, often in minutes. And that fascinates me. I hope it uses AI.

Footnotes

Yao et al., 2022, Princeton/Google Research

The Myth of La Masia vs. La Fábrica

Mon, 14 Sep 2026 00:00:00 GMT

In 2010, FC Barcelona’s La Masia (“The Farmhouse”) became the first football youth academy in history to train all three finalists of the Ballon d’Or in a single year: Lionel Messi, Andrés Iniesta, and Xavi Hernández. Messi won that year, added 7 more to his career, and established himself as one of the greatest players of all time. Beyond individual accolades, La Masia has produced the most World Cup winners. Carles Puyol, Gerard Piqué, and Sergio Busquets were central to Spain winning in 2010, while Pau Cubarsí, Dani Olmo, and Lamine Yamal helped Spain win again in 2026. The latter is also one of the brightest stars of the current time. Another academy product, Pep Guardiola, propagated La Masia’s possession principles across Barcelona, Bayern Munich, and Manchester City, winning every major trophy there is to be won.

Between 2002 and 2018, Barcelona fielded at least one academy graduate in their starting XI for 5,825 consecutive days. Manager Ernesto Valverde broke the streak against Celta Vigo to rest his youth stars for the cup final. Over the past 50 years, Barcelona has won 20 La Liga titles and 5 Champions Leagues while integrating over 550 academy graduates into the first team. Their youth system has shaped their history.

Real Madrid offers a stark contrast. When Madrid won three consecutive Champions League titles between 2016 and 2018, their preferred starting XI featured only one La Fábrica (“The Factory”) graduate: right-back Dani Carvajal. The vast majority of core talent who built the dynasty that won 9 Champions Leagues, the most in history, arrived through the Galácticos recruitment model, a commercial marketing flywheel, that targeted global superstars like Luis Figo, Zinedine Zidane, Ronaldo Nazário, Sergio Ramos, David Beckham, Kaká, Cristiano Ronaldo, Karim Benzema, Gareth Bale, Eden Hazard, Jude Bellingham, and Kylian Mbappé.

Real Madrid has repeatedly fielded teams without a single academy player. In a January 2023 match against Villarreal, they broke a 121-year-old record by fielding a starting XI without a single Spaniard, let alone a youth product. Over the past 50 years, Real Madrid have won 19 La Liga titles and 9 Champions Leagues by spending over €2.9 billion on player transfers. Transfer market dominance, rather than academy integration, has shaped their history.

These parallel histories tempt a simple conclusion: Barcelona builds success, while Real Madrid buys it.

The financial data dismantles this narrative across three dimensions:

Expenditure: In the past 50 years, Barcelona have spent €2.8 billion on the transfer market, a meager 2.5% less than Real Madrid. That nets to €57 million per season, versus the €58 million that Madrid spent.
Asset Monetization: While Barcelona uses its academy talent to fill first-team slots, Real Madrid treats La Fábrica as a high-yield capital asset. Barcelona has generated roughly €250 million selling academy talent; Real Madrid has generated nearly double, €425 million. This model has helped Real Madrid sustain long-term financial stability, consistently ranking as the world’s most valuable football club, while Barcelona continues to navigate severe financial distress.
Talent Footprint: La Fábrica graduates log more top-flight minutes across Europe’s top 5 leagues than any other system. CIES Football Observatory data shows that between 2005 and 2025, 166 La Fábrica products accumulated 1.08 million minutes, surpassing La Masia’s 156 graduates and 1.04 million minutes.

La Masia is not strictly a bountiful farm harvesting homegrown talent, nor is La Fábrica a rudderless factory producing mediocre talent. Instead, the two clubs operate on fundamentally different structural models and measure success through different metrics.

Barcelona views La Masia as an ideological anchor. It instills a uniform tactical language from the U-10s to the first team, allowing young players to seamlessly transition into senior roles. Real Madrid operates on universalism: assemble the best players on earth under one roof using La Fábrica as a high-margin talent engine to generate liquidity needed to fund its Galáctico signings.

Both teams occasionally deviate from these models under strain. In the late 1980s, facing financial constraints and a title drought, Real Madrid turned inward to promote La Quinta del Buitre (“The Vulture’s Cohort”) comprising Emilio Butragueño, Míchel, Manolo Sanchís, Martín Vázquez, and Miguel Pardeza, who led the club to five consecutive La Liga titles and 2 UEFA Cups. Conversely, in 2014, Barcelona bypassed La Masia graduates to spend €180 million on Neymar and Suárez, forming an attacking trio with Messi that secured the 2015 treble.

Yet both institutions inevitably return to their core DNA. Today, La Masia graduates like Cubarsí, Olmo, and Yamal form the tactical spine of Barcelona’s first team, while La Fábrica sales have generated over €200 million this transfer window to subsidize Madrid’s latest superstar additions. One offers fans a mirror to see themselves in the squad; the other offers a telescope to gaze at the stars. And I prefer the latter.

Posers

Mon, 07 Sep 2026 00:00:00 GMT

Birds are my favorite subjects to photograph. It is probably the millennial in me but the feeling you get when you capture them staring into the lens is unmatched. Take a look at some of those moments from Kabini National Forest in Karnataka, India.

The Leak in the Moat

Mon, 31 Aug 2026 00:00:00 GMT

A few weeks ago, prior to writing “Three Salesmen, No Connector,” I was actively reading The Tipping Point—highlighting and taking notes—switching back and forth between my Kindle and the keyboard on my desk. In retrospect, this was a painful experience. Casual learning in the digital world should be much simpler. I just need a device containing my books and notes, and my Kindle Paperwhite technically stores both. Why do I then only use it for its moat, the core reading experience?

The Kindle provides a sub-par note-taking functionality. While entry is relatively easy, the software is slow, organization is overwhelming, and retrieval is hard. So, whenever I read on the Kindle, I always have another device, either my iPad or a Mac, to take notes in Obsidian or Bear. The constant switching between two devices has led me to use my Kindle much less. I don’t get enough value out of it compared to my other hardware and have considerably slowed my upgrade cycle.

Products, especially category leaders from incumbents, over-index on creating a world-class experience for the primary functionality while neglecting supplementary experiences. This creates friction for users, leading them to find alternative solutions for that secondary job. The incumbents will continue to retain users because of prohibitive switching costs, but will lose these secondary jobs to other products. Users will spend less time on the incumbent’s products and affinity will go down, becoming an invisible leak in what would appear to be a healthy funnel only until it’s no longer invisible.

Instagram presents another example. Not too long ago, Meta was not built for vertical video. TikTok capitalized on this opportunity by making vertical video creation, distribution, and consumption extremely easy. It began capturing the audience’s attention—the same audience that engaged with content on Instagram. But these users didn’t delete their Instagram accounts. They just devoted less time there. This quickly turned into an existential risk for the market leader, culminating in the launch of Reels and the shift to a video-first strategy (much to my chagrin, but that’s a whole other post).

Incumbents wedge segments to serve users with secondary jobs. Amazon created the Kindle Scribe for note-takers, and Meta (temporarily) created Lasso for vertical videos. The failure of these products and the rise of products like reMarkable and TikTok show how quickly the invisible leak can become visible.

Ben Thompson’s Aggregation Theory holds that internet-native platforms can serve every user near-zero marginal cost. Their primary investment comes in the form of huge fixed costs, but once the solution is scaled, every additional user can be served without considerable additional spending. And AI-accelerated product development has only driven the cost even lower. But rather than leveraging AI to patch the invisible leak, the incumbents are using it to further strengthen the visible moat. I would fix the leak before the water in the moat dries up.

Possible, Not Plausible

Mon, 24 Aug 2026 00:00:00 GMT

On September 27, 2011, in the 25th minute of Real Madrid’s UEFA Champions League tie against Ajax, Cristiano Ronaldo scored, finishing off a flowing counter-attack involving Káka, Mesut Özil, and Karim Benzema—one of the 174 goals Real would score in all competitions that season. The manager who architected that sublime brand of football 15 years ago: José Mourinho.

When Mourinho was appointed Real Madrid’s coach again in July of this year, proponents of his appointment shared clips of that goal against Ajax citing it as the brand of football Madrid will play again this upcoming season. And, why not? Real Madrid has the talent to play such scintillating, fast-paced football: Mbappé, Vinicius, Diomandé, and Bellingham.

Possible, not plausible.

Reference Class Forecasting

Reference Class Forecasting¹ is when you forecast an outcome by pulling the actual distribution of results from a class of comparable past cases, and placing your case within that distribution, instead of forecasting from the specifics of your own plan.

For example, say you’re estimating the budget for a home improvement project. Reference Class Forecasting says that your estimate based on your idea of what the project needs will be less accurate than an estimate based on costs of similar projects at a large set of similar households.

Reference Class Forecasting involves three distinct steps:

Identify a reference class — a set of truly comparable past cases, large enough to be statistically meaningful.
Establish the outcome distribution across that class — the full spread of how those cases turned out.
Locate the current case within that distribution — accounting for the risk you’re willing to accept.

Step 1 is critical here. The wrong reference class leads to an inaccurate forecast.

What’s the right reference class?

Madridistas set expectations against three reference classes.

1. Mourinho’s last successful season at Real Madrid (2011-12)

The most memorable season that José Mourinho had at Real Madrid was in 2011-12. The club won La Liga claiming a 100 points and scoring 121 goals. It failed to reach the Champions League final via penalties. It is natural for a fan to want that back. But, unfortunately, it’s an invalid reference class, one that is based on salience—wherein the most memorable (or generally the most cognitively available) class is used for comparison instead of the most causally similar one. A lot has changed in the past 15 years: Ronaldo, Guardiola’s Barcelona, and Bayern have been replaced by Mbappé, Flick’s Barcelona, and PSG. Mourinho himself hasn’t managed or won any major trophy at a top European club for a few years now.

2. Real Madrid’s last trophy-laden season (2023-24)

So, how about a more recent benchmark like the 2024 Champions League and La Liga winning team? Can this team emulate that one? It still has most of the stars from two years ago and has since added a whole host of stars including Mbappé, Trent, Diomandé, Bernardo, Cucurella, Dumfries, Konate, and Carlos Espí, a more agile Joselu. Unfortunately, this reference class fails too, this time on the illusion of continuity where shared personnel gets mistaken for shared circumstances. One of the primary reasons for Madrid’s agony over the past couple of seasons has been the missing Connector. That has not been solved yet and a solution isn’t imminent either. And even if Madrid signs one now, it’d take the player some time to integrate into the squad.

3. Real Madrid’s last season (2025-26)

This is the valid reference class. The best way to forecast what Madrid can achieve in the upcoming season is by seeing how much it has improved over last season from the standpoint of personnel and performance. The squad has clearly been upgraded in several positions:

Bernardo Silva over a jaded Dani Ceballos
Ibrahima Konate over an aging David Alaba
Marc Cucurella over a mediocre Fran Garcia
Yan Diomandé over a raw Franco Mastantuono
Denzel Dumfries over an aging Dani Carvajal But the questions below remain:
Who will play Connector? With the non-signing of Rodri, and the signing of Diomandé, Madrid is now left with, not three, but four Salesmen and no Connector.
Can Valverde and Tchouameni co-exist? If this is the preferred double-pivot, can they play together again?
Where will Güler, who’s had a stellar pre-season so far, play?
Carreras, Huijsen, Rudiger, and Konate² are all coming off a terrible season. Is their regression a dip or is it permanent? What role will Joan Martinez, who looked great in the friendlies, play?
Endrick confirmed he’s staying at least until January. What role will he play given Espí has signed?

It is easy to form an Inside View³ here. The defense has been fortified. Mourinho will (magically) mold one of Valverde, Güler, or Bernardo into playing Connector, complementing Tchouameni or Camavinga in the double pivot, and supply the four Salesmen: Bellingham, Diomandé, Mbappé, and Vinicius, challenging Barcelona for La Liga and PSG for the Champions League. This is optimism bias and it requires a combination of a lot going right for Real Madrid and a lot of going wrong for its competitors to work. Instead, try taking the more accurate Outside View⁴. All these unknown unknowns cannot be brushed under the “Mourinho will solve it” carpet. Carlo Ancelotti, Xabi Alonso, and Alvaro Arbeloa—all individually brilliant managers—collectively failed. Mourinho needs time to resolve these uncertainties, and patience is not the club hierarchy’s virtue.

(Realistic) Expectations for this upcoming season

So, what should fans realistically expect from Real Madrid?

Identify a spine and back it: This is the Connector problem, but amplified. Real Madrid has a wobbly spine⁵ on top of missing the Connector. Barring Thibaut Courtois and Kylian Mbappé, the identity of the rest of the spine is still uncertain. Finding that identity and giving that spine the time to brew will set Real Madrid up for years to come.
Be more competitive in big games: Last year’s failures against Atletico, Barcelona, Liverpool, and Bayern Munich were telling. La Liga effectively gets decided by how well the top 3 do in their games with each other. Champions League fate is decided by how a team does against the likes of Arsenal, Bayern Munich, or Paris Saint-Germain. Madrid’s progress will be rated based on how competitive the team is in the biggest games.
Less drama: Last season was fraught with drama—mostly involving Vinicius, Mbappé, Valverde, Tchouameni, Rudiger, and Carreras. This kind of drama has not happened at Madrid for several years now. Mourinho has the habit of bringing the spotlight on himself so that it stays away from the players. Madrid needs that from the manager more than ever.

I want a trophy too: Mourinho after all won the Copa del Rey in his first season in his previous stint. But wishing for a repeat of that feat is salient and would be an invalid reference class. That said, if this team shows me that it has an identity, can stick together, compete in big games, and plays the same brand of scintillating football as it did that evening against Ajax on September 27, 2011, then it will be a season to remember and my reference class for the next.

Footnotes

Reference Class Forecasting as a three-step forecasting method was coined by Flyvbjerg (2006). Its roots trace back to “Planning Fallacy” (Kahneman & Tversky, 1979), and “Inside View vs. Outside View” (Kahneman & Lovallo, 1993).
At Liverpool mostly, but also with France at the World Cup.
This is when you focus on the specific case, build a mental model of how it will unfold, and forecast by extrapolating that model forward — this task, then that one, allowing for the obstacles you can currently foresee. This is how essentially everyone naturally plans, and it’s not stupid — it uses all the detail you actually have about your situation.
This is where you deliberately set the specifics aside, treat the case as one member of a class of similar cases, and forecast using how that class has actually performed, not how this instance is supposed to go.
Goalkeeper -> central defense -> central midfield -> center forward

Chennai

Mon, 17 Aug 2026 00:00:00 GMT

Chennai was my home for nearly 25 years, and it will remain my first home for the rest of my life. It only makes sense to start here. These photos capture a few iconic spots across the city, from Kapaleeswarar Temple to the old TV station.

The Fast Fashion of Code

Mon, 10 Aug 2026 00:00:00 GMT

Vibe coding is the process of developing software with AI by describing in natural language what to build and letting the model do the technical work required. Coined by Andrej Karpathy, it has risen in popularity, making software development accessible to anyone with an internet connection and an AI chatbot subscription. To see this in action, consider a real-life example of an app I recently vibe coded. As a creator, I post my photos on Instagram. I plan my posts in advance, including what to post, when to post it, what features to highlight, descriptions, hashtags, and other metadata. As a visual thinker who likes to keep things organized, I wanted an easy way to accomplish two goals:

A project management tool: A workspace tailored specifically for Instagram posts to help me plan and store all metadata for past and future posts.
A visual grid planner: A interface to preview my Instagram profile, allowing me to try different arrangements of posts and story highlights so I can see exactly what my feed looks like before publishing.

I did a cursory search to see if any free, readily available tools fit the bill. All I found were freemium, web-based applications. I preferred a Mac-native app that kept my data stored locally. This felt like the perfect opportunity to vibe code, as I had a specific, personal, persistent need that I could articulate clearly. After all, software development is what I do for a living, and I have computer engineering experience programming apps.

Using plain English, I articulated the software I wanted to Gemini, and it wrote the code to create an app with the exact user interface, features, and privacy protections I requested. In a matter of a few hours, I had a tailored tool that I now use daily to plan and visualize my Instagram posts.

MKT Studios Project Management Tool v2.3

The Prosumer Breakthrough

I am by no means an early adopter of vibe coding. Andreessen Horowitz ranked consumer and prosumer vibe coding tools, such as Lovable and Replit, as the breakout productivity trend of 2025—and this wave shows no signs of slowing down.

The prosumer gains are undeniable. AI tools are significantly benefiting software professionals. According to a 2025 Google survey, 90% of respondents reported using AI tools at work—up 14% from the prior year—with over 80% noting that AI made them more productive. Anecdotally, I agree. I use AI at work for research, brainstorming, and prototyping. It has made me noticeably more productive, allowing me to spend more time thinking and less time executing.

In theory, AI proliferation should translate to higher value for both the enterprise and end consumers. But how will vibe coding influence consumer software at scale?

Hyperpersonalized Software

The single biggest advantage of building my own app over buying a mass-market alternative was hyperpersonalization. The app I vibe coded was custom-fitted to my exact workflow and aesthetic preferences. I no longer had to settle for software simply because its feature set checked most of my boxes. I just needed to know what to ask for and how to ask it.

Is hyperpersonalization enough to help vibe coding cross the chasm into becoming the primary way consumers use software? Will build ultimately prevail over buy? Consider an analogy: almost everyone would look better in custom-fitted garments. However, fast fashion became the default attire because of mass-production cost efficiencies and a scarcity of affordable bespoke tailors. Industrialization improved affordability, making off-the-rack clothing standard.

Mass-market software thrived for similar reasons: the high fixed costs of software infrastructure and the specialized skills required to program applications meant scale and standardized user experiences were the logical paths forward. Can vibe coding eliminate these historical barriers?

Cost Efficiencies

Vibe coding effectively makes front-end code generation free, but underlying infrastructure remains expensive. The moment a hyperpersonalized app requires cloud synchronization, push notifications, or API authentication, the user stops being a casual creator and is forced into the role of a DevOps engineer—reintroducing the technical barrier. While platforms like Lovable and Replit aim to abstract this friction, users still face real constraints: recurring compute costs and the risks of delegated security permissions.

Specialized Skills

Vibe coding shifts the skill sets required to build software. You no longer need to master syntax, complex architecture, or design patterns. Instead, success requires:

An underlying, well-defined need
The ability to articulate that need clearly
The willingness to spend time building and testing
Patience to QA, iterate, and refine

The persistence to maintain and update the software over time Skills will remain a meaningful barrier. Humans are not naturally adept at articulating precise technical constraints. We are impatient by nature, and we prefer to spend our limited energy on outcomes rather than maintenance. A mainstream consumer must be highly motivated to vibe code a hyperpersonalized solution from scratch.

The Opportunity for Mass-Market Software

The popularity of vibe coding and agentic engineering presents a massive opportunity for scaled, mass-market applications: ephemeral experiences.

As the cost of generating UI on the fly approaches zero, scaled apps no longer need to rely solely on rigid, one-size-fits-all features. Instead, they can generate ephemeral, hyperpersonalized experiences dynamically. Mass-market software will thrive by pairing robust backend infrastructure with a generative frontend customized to each user. Imagine opening your favorite app and having its interface instantly structure itself around your exact intent.

Vibe coding is here to stay, but so are mass-market applications. What will continue to evolve are our needs and expectations, and the solutions that adapt most seamlessly are the ones that will survive and thrive. And for those who haven’t tried building your own software yet, I strongly recommend it. It is engaging, empowering, and deeply satisfying. There has never been a better time to build.

Three Salesmen, No Connector

Mon, 03 Aug 2026 00:00:00 GMT

On June 3, 2024, two days after winning its 15th Champions League title, Real Madrid announced the highly anticipated signing of Kylian Mbappé. The French world champion and arguably the most talented footballer on the planet arrived on a free transfer from PSG. This was yet another massive coup for the inimitable Florentino Perez. And in the eyes of Madridistas, Real Madrid had just won football.

Now, two years later, the tables have turned. PSG has since claimed back-to-back Champions League titles with prolific attacking football, while Madrid is left with a staggering surplus of talent and a glaring deficit of major silverware.

How is that possible? Isn’t 1 + 1 = 2?

In football math, it clearly isn’t.

The Tipping Point

In his book, The Tipping Point, Malcolm Gladwell theorizes that social epidemics are driven by distinct, underlying rules. Two of these are vital to understanding modern squad building:

The Law of the Few: A tiny fraction of people do the vast amount of work in spreading an epidemic. These exceptional individuals fall into three profiles:

Mavens: People who thrive on knowledge accumulation and are intrinsically motivated to share the raw message.

Connectors: People who possess an unusually large network spanning multiple circles, valuing the bridges that help information spread.

Salesmen: People who can persuade, often through innate, non-verbal abilities, compelling others to believe the message. The Power of Context: The circumstances and spaces we inhabit dictate our actions far more than our intrinsic character does. Environment is everything.

Real Madrid’s plight since signing Mbappé becomes infinitely more explainable when we look at the pitch through this framework.

Applying the Framework

A football club is, in many ways, a microcosm of a complex adaptive system. It thrives on the structural harmony of individuals. To understand why Madrid failed to meet expectations over the past two seasons, we must look at its ecosystem through Gladwell’s Law of the Few and Power of Context.

Take any successful team at its peak in football history, and you will find a delicate balance of Mavens, Connectors, and Salesmen. In this context, Mavens are the managers who architect the raw tactical ideas (think Pep Guardiola or Sir Alex Ferguson). Connectors are the midfield generals who orchestrate the geometry on the field (think Xavi or Roy Keane). Salesmen are the superstars who execute the extraordinary, persuading the players around them to elevate their level (think Lionel Messi or Eric Cantona).

Not too long ago, under Carlo Ancelotti, Madrid struck this balance perfectly. In Toni Kroos, it had the ultimate Connector—a maestro with a supreme read on the game, a geometric understanding of the players around him, and the know-how to enable his Salesmen. He was set up to succeed by a balanced front line: Vinícius on the left, Benzema (and later Bellingham) in the middle, and Federico Valverde or Rodrygo on the right. With Luka Modrić by his side for occasions requiring more dynamism, Kroos orchestrated Madrid’s Champions League wins in 2022 and 2024.

Then, Toni Kroos retired. Real Madrid had lost its Connector.

Mbappé arrived that same summer. Suddenly, Real Madrid boasted three generational Salesmen: Vinícius, Bellingham, and Mbappé.

Vinícius excelled when given free rein on his favorite left wing, isolating defenders on the touchline and using his unpredictable footwork to either round the keeper or fire in from range. Bellingham was at his best when he had the room to make late, darting runs into the box, playing off complementary forwards who didn’t occupy his channels. Mbappé was a hybrid of the two. He was magnetically attracted to the left flank but preferred to make darting, central runs behind the defense. This dynamic—Mbappé vying for the exact same space as Vinícius and the exact same runs as Bellingham—created a profoundly sub-optimal spatial environment.

What complicates this geometry even further is Madrid’s glaring weakness on the opposite flank. Until two years ago, this was solved by Valverde playing on the right wing. This balance across the front spread opposition defenses and created the necessary space for Vinícius and Bellingham to operate. However, that right-sided threat has completely faded, with Valverde taking a central midfield role to cover Modrić’s departure, combined with Rodrygo’s injuries and reluctance to hug the touchline. Without a genuine right winger to stretch the pitch, Madrid has become entirely predictable. Because there is no threat on the right, opposing defenses simply tilt their compact blocks to the left, further suffocating three elite Salesmen who are already stepping on each other’s toes.

Looking Ahead

Real Madrid must solve its two big problems going into the new season.

Balance the Attack

Let’s assume for a second that Mourinho will continue to deploy his preferred 4-2-3-1 / 4-3-3 formation. To fix the Power of Context and draw defensive gravity away from Mbappé and Vinícius, the team desperately needs a reliable node on the right side of the pitch. Theoretically speaking, there are several players in the current squad who can play on the right: Rodrygo, Brahim Diaz, Arda Güler, Endrick, and Franco Mastantuono. The latter three, notably, are left-footed. Then, there’s Valverde, who has done great every time he plays there, and Bernardo Silva, who has played there at Manchester City. And if push comes to shove, both Denzel Dumfries and Trent Alexander-Arnold can be genuine attacking threats too. On paper, plugging a player into this space sounds like an easy solve.

However, the reality of a complex system is different. Rodrygo is recovering from an ACL tear, Brahim has never elevated his status beyond a mildly effective substitute, Güler lacks the explosive pace, Endrick is still raw, and Mastantuono is just not ready for elite-tier European football. Valverde is said to prefer the midfield, and Bernardo Silva is certainly past the age where he can thrive as a wide Salesman in a fast-paced attack. And given the club prefers to rotate its natural fullbacks in the defensive line, it leaves management with no option but to look outward—which is clearly what they are doing, considering the strong links to Yan Diomandé. Is Diomandé the solution to this broken geometry? It is hard to say. At 19, he is an explosive winger who primarily plays on the right but can also play on the left. He is a dribbler, making over four successful dribbles per 90 minutes—a Bundesliga record for last season. While he hasn’t been tested in a major competition for club or country, a recent profile by The Athletic suggests that he is similar to Vinícius but with a better disposition to help the defense.

Personally, I think Endrick deserves another chance to establish this right-sided node. He is 20, left-footed, explosive, and fundamentally shifts the team’s spatial context compared to Mbappé and Vinícius. Compared to Diomandé, statistically, his attacking in a similar league was marginally better at a much weaker team in Lyon despite having joined them only in January. And of course, the club could potentially save over €100 million in transfer budget.

Find a Connector

Stretching the pitch is irrelevant if the ball never arrives. If there’s one area on the field where Real Madrid has to spend money, it would be in midfield to fill the Connector void left by Toni Kroos.

There is no in-house replacement. Güler needs more time, and all the other midfielders currently in the ranks are of a different mold. There are no cheap alternatives. The market is inflated. And not finding a true Connector will very likely continue to hurt the club’s entire ecosystem. Rodri could be the answer. He matches Kroos in his passing ability¹ and has proven himself at the highest level. As a Spaniard, he is familiar with Madrid’s nuances and can be the leader that this disjointed team desperately needs. The downside is probably his age. Kroos came to Madrid at 24 and led the club to five Champions League titles in 10 seasons. While Rodri, aged 30, will not have the same long-term impact on this club, he would very likely have an immediate impact akin to what Mourinho is expected to create.

Madrid should not stop there. It must go out now and find a successor for Rodri, sign him up, and integrate him into the squad so that the next time the club loses its Connector, the replacement is ready to step up.

Conclusion

These are not entirely novel insights—the void left by Toni Kroos has been debated by pundits around the world. But Gladwell’s Tipping Point framework brings structural clarity to the madness Madridistas have witnessed on the pitch over the past two seasons. For the first time since Mbappé’s arrival, this transfer window feels like the club is finally close to solving the puzzle, leaving me with genuine hope and expectation. Here’s to the 16th. ¡Hala Madrid!

Footnotes

Although I’d have to see him do this against Barcelona, this against Juventus, and that against Bayern Munich to believe he’s just as good.