Menu #5 - On‐Chain Horizons: From TAO‐Powered AI to the Future of Money

TAO
Psychology Of Money
Web3 Dinner Club: April (LDN)
The Future of Money
Novel Labs
Brand spice
Chefs Note

This is The Menu.

The UK Web3 operator’s weekly briefing —
what founders, investors, and builders are actually discussing behind closed doors.

This week the theme is:

Three seemingly separate threads:

Bittensor’s incentive-coordinated AI, Morgan Housel’s behavioural truths, and the coming wave of tokenised stocks.

Actually converge on the same question:
What will value and growth even mean in 2026?

In this issue:

• TAO and TAO subnets
• A founder lesson from psychology of money
• Our thoughts on the future of money: How does this all stack up
• Builders as the new growth engine
• Four Pillars Research Report on 2026 being the year of Tokenised Stocks.

Signal, served weekly.

This month’s dinner is proudly sponsored by Novel Labs.

A multi-award-winning London storytelling studio building the brands of the future in AI, blockchain, and emerging technologies.

Best known for the $100m expansion to the Bored Ape Yacht Club, The Mutant Cartel World.

If you’re a startup or scale-up building a brand and looking for real go-to-market impact from those who have repeatedly built unicorns and category kings as VCs and founders... ask for an intro at the table.

Bittensor is a decentralised network for building and coordinating AI where models compete, collaborate, and get rewarded based on the value they provide.

In simple terms:
• Developers plug AI models into a shared network
• Models are ranked on usefulness (not hype)
• Contributors earn TAO for providing value
• Anyone can access and build on top of it

Why it matters:

It’s one of the first systems where:

AI development becomes open, incentivised, and permissionless

One-line summary:

Bittensor = Open-source AI, coordinated by incentives

TAO isn’t just an AI token, it’s the economic layer powering a growing network of specialised AI markets, known as subnets.

Each subnet is its own mini-economy: focused on a specific task (LLMs, inference, compute, data), competing for attention, capital, and ultimately, TAO emissions.

The more useful a subnet is, the more TAO flows into it.

Under the Dynamic TAO (dTAO) model, every subnet operates its own liquidity pool with a native “alpha” token.

When you stake TAO into a subnet, you receive its alpha token, and the market price of that token determines how much of the network’s daily TAO emissions flow back to that subnet.

In effect, each subnet is a mini‑economy competing on usefulness and capital.

By March 2026, the 128 active subnets had a combined alpha‑token market cap of roughly $1.12 billion (about 27 % of TAO’s market cap) — a sign these aren’t just experiments but increasingly significant markets.

Manifold Labs’ Targon (Subnet 4) recently co‑authored a whitepaper with Intel engineers titled

“Decentralized Compute on Untrusted Hardware Using Intel TDX and Encrypted CVMs”.

The paper outlines a confidential‑computing architecture that pairs Intel’s Trust Domain Extensions (TDX) with NVIDIA’s confidential‑GPU technology to protect AI workloads at the hardware level.

This collaboration published on both Intel’s community blog and Manifold’s site marks the first time a Bittensor subnet has released a formal security blueprint alongside a major chipmaker.

It’s about moving from marketing narratives to deployed infrastructure: providing enterprise‑grade security without relying on centralised cloud providers.

Targon is already one of Bittensor’s highest‑revenue subnets, backed by over 1 500 H200 GPUs and a $10.5 million Series A round.

The Intel collaboration strengthens its pitch to enterprises that need hardware‑level confidentiality before adopting decentralized AI.

Together, these developments show TAO evolving from a single token into a broad index of decentralized AI markets and that the most exciting competition and value creation is happening within subnets.

Subnets are no longer just experiments.
They’re becoming credible building blocks for how AI gets trained, secured, and deployed.

Signal:
TAO is evolving from a token → into an index of decentralised AI markets

And subnets are where the real competition and value creation is happening.

Most personal finance books try to teach you what to do.

Morgan Housel’s The Psychology of Money focuses on the real difficult part:
Why we don’t do it.

The central thesis is simple and sticky: doing well with money has less to do with intelligence and more to do with behaviour, and behaviour is shaped by emotion, experience, and incentives, not spreadsheets.

The book is structured as 19 short chapters/stories, which makes it easy to pick up and actually finish.

Housel’s relatable move is taking finance from a math test into being human. People make wildly different money decisions not because they’re irrational, but because they’ve lived different lives with different reference points.

(His famous “no one is crazy” idea lands because it’s true: our money choices often make sense to us.)

A few themes recur, and they’re useful because they change how you act:

It’s unusually good at turning complex ideas into clean stories, without talking down to the reader.

Being human we want it done for us, so if you want a detailed, step-by-step investing plan, this isn’t that.

It’s more like a mental model upgrade, you’ll still need a practical framework elsewhere.

This is actually the core; you have to do it for yourself as you`ve lived you, so you actually (even if you dont like to admit it) know what works for you!

The Psychology of Money is one of those rare books that makes you calmer and smarter about money simultaneously.

It won’t hand you a perfect strategy, but it will reduce the odds you sabotage a good one, by helping you understand your own behaviour better.

Best for: anyone who earns, saves, invests, or worries (so… everyone).

Most networking events optimise for volume.

Not signal.

Too many conversations that go nowhere.

This is designed differently.

The Web3 Dinner Club is a curated, seated dinner for founders, investors, and operators building in crypto, AI, and frontier tech in the UK.

One table (or two). No noise. Just people worth knowing.

Over the past 3 years:

• 25+ dinners
• 230+ curated members
• Outcomes → hires, partnerships, capital

This isn’t about “meeting people.”

It’s about getting in the right rooms consistently.

We keep this intentionally small.
Not everyone gets a seat.

Proudly Sponsored by Novel Labs.

In our view, the future of money won’t feel like a crypto revolution for most people.

It will feel more like the internet: one day you’ll notice the infrastructure has quietly changed beneath the surface, becoming more invisible, more programmable, and always on, yet daily life simply feels smoother, faster, and for some, more convenient.

1) Your money stays in pounds (or dollars).
The unit you use to measure your life won’t change. What shifts is how it moves, faster, cheaper, more programmable, and increasingly automated.

2) The rails change underneath you.
Many everyday “bank transfers” will quietly migrate to new back-ends: tokenised deposits and regulated stablecoin rails. These will enable faster settlement, cheaper cross-border payments, and simpler payouts. You may never see a blockchain screen as the new infrastructure will sit invisibly inside your familiar bank or fintech app.

3) Markets become more “always on.”
Crypto taught the world to expect 24/7 access. As more assets get tokenised (or mirrored on-chain), consumers will increasingly demand the ability to trade, move, or manage money outside traditional 9–5 hours, even if some underlying systems still have limits.

4) Ownership vs exposure becomes confusing.
You’ll encounter more products that behave like stocks, funds, or indices on-chain, long before you truly “own” the underlying asset in a direct legal sense. The wrapper matters: convenience improves dramatically, but you’ll need clear answers on your rights, custody arrangements, and consumer protections.

5) AI will start spending small amounts for you.
The next big shift isn’t you paying differently it’s software acting on your behalf. Subscriptions give way to true pay-per-use models, while AI agents quietly buy services, data, and compute in the background. It’s incredibly useful, but it introduces a new question every consumer will face: who’s accountable when the agent makes a mistake?

Consumer bottom line:
Money is becoming more programmable and more automated, but the products that win won’t be the ones that feel the most futuristic. They’ll be the ones that still deliver what people have always wanted: real trust, strong protections, easy reversibility, and clear accountability.

Here’s the simplest rule for navigating the consumer era of money:

Follow the players who can make these powerful new rails feel as safe and reliable as the old ones.

A couple of weeks ago we had a quietly interesting AI signal from a funding round!

Advanced Machine Intelligence (AMI), founded by former Meta chief AI scientist Yann LeCun, raised $1.03bn at a $3.5bn pre-money valuation to build AI systems based on reasoning, planning, and “world models.”

The bet is a direct challenge to the dominant LLM storyline. LeCun’s view is that systems built mainly on “predict the next word/pixel” are impressive but still fall short of the kind of common-sense, real-world intelligence needed for autonomy.

AMI’s pitch is that real-world data, from cameras, wearables, and industrial sensors, is messy and noisy. Instead of trying to “generate” every detail, their world models aim to learn abstract representations of reality and make predictions in that representation space.

The practical unlock is action-conditioned prediction: an AI that can ask, “If I do X, what happens next?”, then plan a sequence of actions with safety guardrails.

This is AI aimed less at writing and more at operating: AMI says its early customers are organisations running complex systems, manufacturing, automotive, aerospace, biomedical, and pharma, with consumer applications (like domestic robots) as a longer-term path.

If the LLM era made AI feel like a new interface, world models are a bet on AI as a new control layer.

AMI LABS

The report’s core claim is simple: after stablecoins and tokenised Treasuries, tokenised stocks are the next leg of TradFi moving on-chain, and 2026 could be the year!

The scale gap is the point: global equities are ~$140T (US ~$72T), while tokenised stocks are still under $1B. That’s why the upside looks huge!

One useful clarity the report adds: “tokenised stocks” can mean very different things:

W3DC takeaway: when someone says “stocks on-chain,” ask which model, because on-chain exposure and on-chain ownership are not the same.

The other pressure building fast is hours. TradFi is stretching toward extended trading (24/5), while crypto has been always on from day one.

We love beehiiv; it’s where this newsletter is hosted.

If you would like to start your own newsletter, we think beehiiv is the best tool to use.

If you want some help setting one up just reply to this email and I will be happy to help.

The all-in-one platform that brings together newsletters, websites, and every tool you need to grow and earn.

Until next time