Building AI agentic systems. I think and write about business, organizations, learning, and AI through a multidisciplinary, computational lens. DMs open
2026 | The Meta Year of Personal Brand as a Digital Twin of Each Individual
Life is a process of computing, looking for 'the best solution to maximise utility.
You cannot pre-calculate the optimal life path. The system is too complex, with numerous variables, numerous feedback loops, and numerous emergent properties that only reveal themselves over time. It means that by living, it is the only way to compute and get a result of computation. You cannot optimize your life from outside of it. The process of living IS the calculation.
However, the 'utility' above is usually misconceived as a material term. It actually should be measured or recorded in terms of the depth and width of subjective experiences.
Why?
Here's the crucial complication: computation is not value-neutral. While computational irreducibility means you can't predict which life path is "best," it doesn't mean all experiences are equivalent. Pain and pleasure are not interchangeable data points. Joy and suffering feel different, and this difference is real.
The Recursive Optimization
What's happening, then, is something more subtle than simple optimization:
You're continuously optimizing based on your current values
Your values themselves are being reshaped by your experiences
The "objective function" you're optimizing for is itself a variable in the equation.
This is dynamic optimization with a self-modifying utility function. You can't solve it ahead of time because the very act of solving it changes what you're solving for.
What patterns are emerging in your computation? What parts of your state space remain unexplored? The only way to know is to keep running the program - to keep living, experiencing, computing the answer that only you can calculate.
Sampling is not free product.
It is relationship design.
Most brands think about sampling as a conversion tactic:
Give people the product.
Hope they try it.
Track the redemption.
Calculate the CAC.
That is too narrow.
The deeper value of sampling is that it changes the emotional starting point of the customer relationship.
An ad interrupts.
A sample gives.
An ad says: “Buy this.”
A sample says: “Try this first.”
That difference matters.
If the product is good, sampling creates several compounding effects:
It lowers the first-use barrier.
It gives creators a reason to talk about the product.
It creates social proof from actual experience.
It turns product quality into word of mouth.
It shifts part of the marketing budget from platforms back into the product itself.
Not everything valuable can be cleanly attributed.
The path is messy.
But the relationship started with generosity.
This is why some growth does not show up neatly in attribution dashboards.
It accumulates through experience, memory, trust, and repeated exposure.
For products with low trial friction, sampling is not a side tactic.
It can be the first layer of the growth system.
When AI gets stronger in planning and execution, humans need to move to more abstract thinking, which means higher representational in the hierarchy of perception, a bird's-eye view capturing a larger scale of information in real-time.
Deep thinking and reading prompted deep learning is the only path, I guess.
I think the Lead Quality is Bad or Poor should be thrown into histories soon.
New generation of organisations should think and behave systamatically, rather than at a single surface point.
One lead could be 'bad', but this is not important.
To observe and build a model, is the key.
What it means, to keep that mindset of conversion rate, can help us focus on the signal for a decision, not to conclude based on a subjective experience.
Building a customised CRM for any business should be a trend since 2026.
And what can really differenciate the CRM in the era of AI is how the data visualised in terms of roles and responsibility.
CRM should be a powerful 'dashboard' giving insights, which should be integrated into an agentic system/loop engineering for a better understanding of the business and inproving the organisational collabaration, instead of just recording the data as pipelines which almost irrelevant to human's eyes & brains.
A great brand brief does not describe aesthetics.
It describes strategic tension.
Many brand briefs sound like this:
Make it premium.
Make it modern.
Make it clean.
Make it bold.
Make it feel healthy.
That is not strategy.
That is taste language.
A useful brief creates a problem worth solving.
For example:
How do you signal performance without speaking only to men?
How do you make science feel desirable instead of clinical?
How do you make a functional product feel cultural?
How do you make something modern while giving it symbolic depth?
That is where branding starts to matter.
Because now design is not decoration.
It becomes a decision system.
The name, color, packaging, typography, claims, copy, and campaign system all have to resolve the same strategic tension.
When the brief is weak, design becomes subjective.
Everyone argues about taste.
When the brief is strong, design becomes operational.
Every choice can be judged against the same underlying question:
Does this make the strategy visible?
Strong brands are not built by asking design to “make it look better.”
They are built by translating strategic tension into symbols people can feel.
AI is becoming a category convention amplifier.
It can generate the average of everyone’s homepage, ad copy, and positioning deck.
That makes “obvious and true” content almost free.
The real operator edge is finding what is not obvious but true: the product truth, category gap, or system constraint that has not been named yet.
Another new thing delivered with Version 15: click a button to give Wolfram superpowers to your AI system! Immediately connect Claude, Codex, etc. to your local Wolfram|One, Mathematica, etc. via MCP...
https://t.co/WGtJmebGPa
AI does not lower the bar for brand. It raises it.
The most common mistake I see right now is treating AI as a productivity tool for brand work.
More copy.
More ad variants.
More social posts.
More landing pages.
Faster, cheaper, easier.
This is the wrong frame.
Because if AI lowers the cost of producing average brand content for you, it lowers it for your competitors too.
The result is not advantage.
The result is noise inflation.
Every category gets louder, every feed gets more crowded, and every customer gets more numb.
The honest read is that AI does not lower the bar for brand work.
It raises it.
What gets cheaper is execution: copywriting, design iteration, asset variation, A/B testing, localization.
What does not get cheaper is the upstream work: deciding what is true about your product that nobody else can credibly say.
Choosing which customer you actually want.
Defining the tension your brand resolves.
Building the trust that survives a category-wide content flood.
This is the part AI cannot do for you, because it does not know which truth is yours.
The companies that win the next five years will not be the ones who generate the most.
They will be the ones who decide the most.
Which standard to hold when the rest of the category lowers theirs.
AI is a leverage multiplier on judgment.
If you have judgment, it makes you ten times more effective.
If you do not, it makes you ten times more average, faster.
The question is no longer "are you using AI."
The question is what part of your brand judgment you are still willing to own.
The dashboard cannot see the strongest growth
Every dashboard I have ever seen rewards the same thing: attribution.
Click → conversion.
Channel → revenue.
Spend → ROAS.
So that is what teams optimize for.
But the strongest brand growth I have watched happen up close almost never shows up cleanly on a dashboard.
A sample arrives in someone's mail.
They give it to a friend.
The friend mentions it at the gym three weeks later. Someone overhears it.
Two months after that, an order shows up on Amazon with no traceable source.
Attribution says "direct."
Attribution is lying.
The growth was real.
The chain was real.
The dashboard just could not see it.
This is the part most growth teams refuse to sit with:
a meaningful portion of brand demand is built in paths the analytics stack was never designed to measure.
Not because tracking is broken.
Because the thing being tracked is not a click.
It is a memory.
It is a story someone told someone else.
It is a moment of trust that compounds quietly for months before it converts.
The real question is not "can we attribute it."
The real question is: would we still do it if we could not?
If the answer is no, you do not have a brand strategy.
You have a media plan.
What is one thing your brand does that you cannot attribute, but you keep doing anyway?
After programming, coding agents will boost workflow agents to take off.
But different from coding with clean and historical accumulated structural input/output protocles, workflows are messy, chaotic, unknown, and nobody can articulate the whole picture, as thinking in systems is rare.
However, the world is shaping and 2026 is the Meta year of Agentic workflow adopted by an increasingly large group of businesses.
Cursor CEO Michael Truell on the future of writing code: "Our goal with Cursor is to invent a new type of programming."
"It looks like a world where you have a representation of the logic of your software that does look more like English."
"You can imagine kind of an evolution of programming language towards pseudocode. You have written down the logic of the software, and you can edit that at a high level."
"It won't be the impenetrable millions of lines of code, it'll instead be something that's much terser and easier to understand and easier to navigate."
@mntruell with @lennysan on Lenny's Podcast
The “not obvious, but true” line is the one I keep coming back to.
But the real edge is not saying it.
It is knowing how to find it.
That feels like what separates Peter from most operators.
He does not just make the marketing interesting.
He finds the product truth that is already interesting, then builds the brand around translating it.
Great Content, appreciate Mark
By connecting with Claude Code or Codex, non-coder as me can start to picture myself using Wolfram language to develop a more reliable agentic system adopted in business domains.
Would be helpful if any demo videos for different levels of users to get onbard faster. With the vibe coding rushing into the real world, what the AIs actually computing will become increasingly important soon.
Great LIst! Very helpful.
The automated workflow absolutely 'improve' how we humans 'can' do and track.
The list always seems perfect and ready to use, what really limits the potential is to start using the real basic and build up the automation as small features step by step.
By far, I just used Obsidian as a content scrapper and what really matters is how my brain can memorise and visualise the projects and relevant pieces of information.
What LLM can memorise and process does not equal to what human brains can digest and also maintain a good quality of output.
What feature serves people to SEE better probably a potential service-as-a-software idea. 😂
This is one of the cleanest outbound builds I've seen, and the most important line in it is easy to miss:
"The record in the middle is what makes the whole thing more than a mail merge."
Exactly.
The agents aren't the system.
The shared record is.
Six clever prompts with no memory is just a faster spray.
But here's the next problem nobody hits until they scale this:
once that record holds enough, it stops being only facts. It starts holding the engine's own guesses too.
A score is an inference.
A "why_now" is an inference.
And the moment those get written back in next to the real triggers, your warm account memory quietly fills with things the machine assumed, not things that happened.
That's the layer I wrote about:
what gets to count as truth when an AI both reads and writes the record.
https://t.co/pv2IkH5m3m