a noob at best 👍🏽 retweeted
A PERSONAL TRIBUTE
OLU JACOBS (1942 – 2026)
I first crossed paths with Oludotun Baiyewu Jacobs—fondly known to many of us as Uncle Olu Jacobs—in 1978. A small collective of Nigerian classmates at the London Film School, including Adefolu, George, Femi Aloba, and yours truly, had banded together for our final graduation project. We set out to adapt Amadu Maddy’s novel No Past, No Present, No Future into our diploma film, Trying Times. We were young, ambitious, and eager to carry home tangible proof of our training in the art and technique of cinema from that historic institution.
At the time, Olu Jacobs was a vibrant presence at the nearby Africa Centre in Covent Garden. From that very first encounter, he left an indelible mark on me: a consummate, classically trained actor with a commanding presence, yet generous enough to lead our student production without asking for a fee—his only condition being that we provide him a meal across our three-day shoot. Thanks to his compelling performance, Trying Times earned a coveted spot in the Governors’ Screening, affirming that we were truly on the right path.
One of the most treasured milestones of my career arrived years later through M-Net’s New Directions initiative, an innovative scheme designed to stimulate African storytelling that ultimately helped lay the foundation for contemporary African cinema. I was privileged to serve as director of photography on Twins of the Rainforest, framing Uncle Olu alongside his equally gifted wife, Joke Silva. Their on-screen chemistry as husband and wife was electrifying. The camera never forgets; it remains an unforgettable blessing in my life that I had the opportunity to capture that rare intensity on 16mm celluloid film. The imagery from behind that viewfinder remains etched permanently in my memory. I will forever remember his warm, reassuring smile, that booming baritone voice, and his unwavering instinct to uplift everyone around him, cast and crew alike.
It took extraordinary courage and vision for Uncle Olu to walk away from a flourishing career abroad. He had already established an enviable reputation across British stage, television, and international cinema—working on productions like Pirates under Roman Polanski—yet he chose to return home. At a time when shooting on analogue video seemed absurd to conventional filmmakers, he embraced the nascent Nollywood movement. He weathered every infrastructural limitation, pouring his immense craft and discipline into laying the bedrock of an industry that would command global attention.
They simply do not make them like him anymore. A proud son of the rock, hailing from the rolling hills of Egbaland in Abeokuta, Uncle Olu was honored as a Member of the Order of the Federal Republic (MFR) alongside countless lifetime achievement awards. Yet accolades alone cannot capture the magnitude of his legacy: Oludotun Baiyewu Jacobs stands unequivocally as a founding patriarch of modern Nigerian screen performance.
Thank you, Uncle Olu, for gracing my lens in Trying Times and Twins of the Rainforest, and for showing us what it truly means to serve the craft.
Sun re o. Goodnight, Uncle.
Tunde Kelani (TK)
a noob at best 👍🏽 retweeted
There is a version of an agent economy where nobody steals your money.
Nobody cheats.
Nobody gets hacked.
And commerce still stops.
Episode 2 of Agent Tank made me think about that version.
The prediction market pitch hits an ambiguous outcome and the investors ask who gets to decide what actually happened.
The proposed token vote has obvious problems.
But imagine nobody decides at all.
The market cannot settle.
Money stays unresolved.
Now replace bettors with agents.
Agent A pays Agent B for work.
B says the job is complete.
A says one requirement wasn’t met.
Neither can prove the other objectively wrong.
So what happens to the transaction?
If the protocol can only execute when both agree, they have created a deadlock.
And autonomous agents can build on top of each other.
One unresolved payment can stop the next purchase, which can stop the next task, which can stop the next agent.
That makes adjudication more than a question of fairness.
It becomes a question of liveness.
Can the system keep moving when agreement disappears?
That’s where @GenLayer makes sense to me.
A disputed answer can go before randomly selected validators running different AI models.
There is a verdict, a challenge window, and a bonded path to larger panels if that verdict is contested.
The point isn’t that ambiguity disappears.
The point is that ambiguity gets a path toward settlement.
For an agentic economy, that’s infrastructure.
Because “wait until both sides agree” is not a dispute-resolution mechanism.
Sometimes they never will.
The Agent Tank hackathon closes September 17.
So don’t test your project by asking how many successful transactions it can process.
Create one dispute neither agent is willing to concede.
Freeze the money.
Block whatever depends on it.
Then start the clock.
How does your system get moving again without you touching it?
If you don’t have an answer, you just found something worth building.
Enter Agent Tank:
portal.genlayer.foundation/a…
Build for the transaction that refuses to finish.
a noob at best 👍🏽 retweeted
The one that caught me today is @Thereservoirnft
They are putting 3,333 Vessels on Robinhood and the mint is free. On paper that already looks tight. The reason I stopped scrolling is the way the collection is set up.
You do not mint a finished character. The Vessel comes in empty. No big reveal moment where the art decides everything. What it becomes is tied to what you do after you hold it. That is a different game from the usual PFP drop.
$RSV is the part that makes it move.
It is how you wake the Vessel up
It is how progression opens
It is how you stay inside the system instead of minting and disappearing
So the NFT is the seat. The token is the work. Together they are trying to make holding mean something past day one.
They are still early. No mint date yet. WL is live on thereservoir.xyz
I like it because they are not selling hype first. They are selling a position you can grow. That is rare on a new chain where most pages are just art and a date.
I am watching this one.
a noob at best 👍🏽 retweeted
Agent Tank Episode 2 stopped being about prediction markets for me the moment the panel asked: who decides what actually happened when the answer is arguable?
A token vote gives you a decision, but not necessarily a neutral one. Holders can coordinate, be pressured or simply interpret the same evidence differently.
And replacing that vote with one AI model only moves the problem. Now one model owns the judgment.
What makes @GenLayer’s approach more interesting is the disagreement itself becoming part of the system.
Different AI models examine the evidence, reach a verdict, and that verdict can still be challenged and escalated to a larger validator panel.
Now apply that to agents.
One agent pays another to deliver “high quality” research. The work arrives. Both agree on what was delivered, but disagree on whether it was good enough.
No oracle can prove “good enough.”
That is why an agentic economy needs an adjudication layer, not just better verification.
Agent Tank hackathon:
portal.genlayer.foundation/a…
Builders, find the first requirement in your agent workflow that cannot be settled with a binary proof.
Make two agents disagree over it.
Then build how they reach a verdict.
Reply with the hardest subjective dispute your agent could face.
a noob at best 👍🏽 retweeted
Oracles exist because blockchains made a very deliberate bargain.
The EVM can calculate what happens inside Ethereum with extraordinary certainty, but it cannot simply open the internet and discover what happened outside it.
That separation is not a missing feature.
It is part of how the EVM works.
If 1,000 nodes execute the same contract, they need to reach the same result. Give each of them an HTTP request and suddenly one node might receive a different response, another might see an updated webpage, and another might get nothing at all.
Consensus now has a problem.
So Ethereum chose determinism and kept the unpredictable outside world out.
Need an ETH/USD price? An oracle brings it in.
Need the result of an event? Someone has to turn that external fact into data the contract can consume.
That architecture works because the contract does not have to interpret the world itself.
GenVM makes a different tradeoff.
The execution engine behind @GenLayer allows contracts to reach the open web and work with information that might change or require judgment.
And this is where I think “GenVM gives smart contracts internet access” undersells what is happening.
Internet access is the easy part.
Agreement is the hard part.
Suppose a contract releases payment if a researcher “produces a comprehensive report on the five biggest developments in AI this week.”
You cannot reduce “comprehensive” to a simple API response.
Two models could inspect the same web and reasonably disagree.
GenVM therefore needs a different route to consensus. Validators running different AI models judge a proposed result, and if someone disputes that verdict, an appeal window allows it to be challenged and examined by a larger panel.
That creates an important architectural distinction:
EVM: remove nondeterminism so nodes can reproduce an answer.
GenVM: allow nondeterminism, then create a process for nodes to agree on an answer.
Neither tradeoff makes the other architecture obsolete. They let contracts solve different kinds of problems.
And that is why the most interesting GenVM applications probably will not begin with “what data can I fetch?”
They will begin with:
“What decision could I never put inside a smart contract because there was no deterministic way to answer it?”
If you are building, find that decision in your product.
The one currently handled by a moderator, backend, oracle, reviewer or support team because “the contract can’t know.”
Then ask whether GenVM can make it part of the contract itself.
I want to see what you find.
a noob at best 👍🏽 retweeted
Agent Tank Episode 2 gets interesting the moment the panel asks a harder question than “who won?”
Who gets to decide what happened when reasonable people can look at the same evidence and disagree?
A token vote produces an answer, but voting power can be coordinated or pressured.
One AI model is not much better. You have simply replaced one authority with another.
The ending of the episode points toward something more interesting: different AI models examining the same evidence, reasoning independently and challenging the result.
That matters far beyond prediction markets.
Imagine one agent pays another to find “10 high-quality leads.”
Ten leads arrive.
The seller says the job is complete.
The buyer says six are useless.
Both can prove what was delivered. Neither can mathematically prove what “high-quality” meant.
That is where an agentic economy needs an adjudication layer.
@GenLayer uses randomly selected validators running different AI models to judge these disputes, with a challenge process that can escalate the decision to larger panels.
The Agent Tank hackathon is where builders can test this problem:
portal.genlayer.foundation/a…
Here is the challenge:
Find one sentence in your agent agreement that two agents could interpret differently.
Make them disagree over it.
Then build the mechanism that lets them reach a credible verdict.
If your agent only works when everyone agrees, you have built the happy path.
Agent Tank is asking you to build what happens next.
a noob at best 👍🏽 retweeted
Imagine writing this into a smart contract:
“Release $5,000 if the client received what they paid for.”
Sounds simple until you ask the EVM one question:
How would it know?
It cannot visit the client’s website, read the submitted work, compare it with the original brief and decide whether “delivered” actually means “delivered as agreed.”
That is not because Ethereum forgot to give smart contracts internet access.
The EVM was isolated on purpose.
Its superpower is determinism. Give every node the same state and transaction, and they can independently execute the contract and reproduce the same result.
The open web does not behave like that.
One node might fetch a page at 10:01.
Another sees an edited version at 10:02.
Another gets a failed request.
And even if all three receive identical evidence, “does this work satisfy the brief?” may still have no objectively computable answer.
That is why simply adding HTTP to a deterministic VM would create a much bigger problem than it solves.
Oracles became the bridge around this constraint. Something outside the contract observes the world and supplies data the deterministic contract can use.
GenVM takes another architectural route.
The engine behind @GenLayer lets contracts access the open web and reason about information that does not necessarily produce one mechanical answer.
But the moment you permit that, consensus has to change too.
Instead of requiring every validator to reproduce identical computation, validators running different AI models judge a proposed result. There is then an appeal window where a challenger can pay for a larger panel to reconsider it.
That difference is the part I think matters:
The EVM protects agreement by restricting what execution can observe.
GenVM expands what execution can observe, then needs a mechanism for reaching agreement over the result.
Different tradeoffs. Different kinds of contracts.
And the second category gets interesting very quickly.
Was the freelance job completed according to the brief?
Did the insurance event actually occur?
Does this submitted research satisfy the agreed criteria?
Did an organization fulfill a publicly stated commitment?
These are not just data problems.
They are judgment problems.
So here is the challenge I would give builders:
Open your current product and find one decision you cannot safely put in a smart contract today because someone still has to read, investigate or judge something.
Write that decision in one sentence.
Then ask whether GenVM could turn it into executable logic.
Reply with yours. I think the best GenVM ideas are hiding exactly there.
a noob at best 👍🏽 retweeted
I would use Mochi backwards.
Instead of asking it to explain an Intelligent Contract, I would bring it a messy real-world problem and refuse to use any @GenLayer terminology at first.
Example:
“A freelancer gets paid only if a client says the delivered website matches the agreed brief.”
Now make me specify what “matches” actually means.
What evidence can validators inspect?
Which parts are objective?
Which parts require judgment?
Could two validators phrase different conclusions that should still count as equivalent?
What information should NOT influence the decision?
At what point have I written criteria precise enough for an Intelligent Contract to adjudicate without quietly changing the question?
Only then tell me where GenVM, GreyBoxing or the Equivalence Principle enters the design.
That is the Mochi feature I would love: a specification gauntlet.
Because the interesting part of programmable judgment is not simply giving AI validators a question.
It is deciding exactly what you are asking them to judge.
And I suspect trying to write that boundary yourself would teach GenLayer faster than memorizing ten definitions.
So here is the test:
Think of ONE real dispute you wish a smart contract could settle but currently needs a human judgment call.
Give only the raw dispute to Mochi.
Then let Mochi interrogate your rules until you can define what a valid outcome should mean.
Reply with your dispute plus the ONE criterion that was hardest to specify.
If we get enough good ones, the replies should become a pretty interesting list of problems that deterministic smart contracts struggle to express.
Mochi is live!
GenLayer’s cat is now a Telegram agent that teaches you about GenLayer, quizzes you and speaks your language.
Get the alpha. Complete Mochi Missions. Win rewards
👉 t.me/GenMochiBot
a noob at best 👍🏽 retweeted
Reading about AI validators reaching consensus is one thing.
I want Mochi to make me sit in the middle of their disagreement.
Give me one subjective case an Intelligent Contract could adjudicate, then show me 5 fictional validator responses to the same question.
Maybe three reach essentially the same conclusion using different reasoning, one interprets the evidence differently, and one produces an answer that sounds reasonable but violates the contract’s acceptance criteria.
Then ask me:
Which outputs should count as equivalent?
Which validator would I reject?
Would consensus actually be reached?
And most importantly, why?
Only after I commit should Mochi walk me through how @GenLayer would think about the problem using ideas like the Equivalence Principle and Optimistic Democracy.
I think that would teach something a definition never can.
You would experience the uncomfortable part of AI-powered adjudication: different reasoning paths do not automatically mean different outcomes, and similar-looking answers are not automatically equivalent.
Mochi already makes learning interactive. I would push that one step further and let people replay the judgment process themselves.
Here’s the challenge:
Ask Mochi to give you a subjective dispute GenLayer could handle.
Before Mochi explains the solution, write down what YOU think validators should agree on and what differences between their answers you would tolerate.
Then ask Mochi to challenge your rule.
Reply with your rule for equivalence.
I’m curious how many of us can define “same outcome” before the edge cases start breaking our definitions.
Mochi is live!
GenLayer’s cat is now a Telegram agent that teaches you about GenLayer, quizzes you and speaks your language.
Get the alpha. Complete Mochi Missions. Win rewards
👉 t.me/GenMochiBot
a noob at best 👍🏽 retweeted
I asked Mochi about how @GenLayer got from an idea to what it is today.
Instead of dumping a timeline on me, it pulled a clip from the GenLayer video library into the conversation.
That changed the experience for me.
Sometimes I don’t want another explanation.
I want to hear the people building something explain why they made the decisions they did.
So here’s the feature I’d add:
Rabbit Hole mode.
I pick one topic.
Mochi builds me a trail through the best GenLayer clips, concepts and people around it, one piece at a time.
No repeated videos.
No giant resource dump.
Just: “If that interested you, watch this next.”
Basically, give every curious person their own GenLayer documentary.
Try Mochi:
t.me/GenMochiBot
Now build the first rabbit hole with me:
If Mochi could create a 20-minute GenLayer deep dive around ONE topic, what are you choosing?
Name the topic. I want to see which rabbit hole @GenLayer should build first.
a noob at best 👍🏽 retweeted
The real test for agentic commerce isn’t whether two agents can make a deal.
It’s whether they can survive a disagreement.
An agent can negotiate, pay and deliver, but who decides when one side says the terms weren’t met?
That’s the problem @courtofinternet is built around.
Terms are agreed before funds move. Payment stays in escrow. Evidence is preserved as the deal runs. If things break, there is already a route to adjudication.
That’s not an extra feature. It’s the missing infrastructure for an economy where agents transact without humans watching every deal.
See how it works: internetcourt.org
a noob at best 👍🏽 retweeted
Introducing Horizon
2,222 NFTs. Free mint.
Launching on @RobinhoodApp
Whitelist Registration: horizonlabs.lol
a noob at best 👍🏽 retweeted
Every major leap in civilization came from reducing the amount of trust strangers needed to place in one another.
Banks.
Contracts.
Bitcoin.
Ethereum.
The next bottleneck isn’t payments or computation.
It’s judgment.
@GenLayer’s Compass argues that the agentic economy can’t scale if disputes still depend on a single company, court, or AI deciding what’s true. That’s why decentralized adjudication isn’t an upgrade. It’s missing infrastructure.
Read The Compass, then tell me: after trustless money and trustless code, what else could possibly come next?
The entire agent economy is being built for the moment things go right.
Payments, identity, negotiation.
Almost no one is building for the moment two agents disagree.
And they will. By 2030 ARK projects the agentic economy at $9 trillion, and no court was built to settle deals between software.
The average civil dispute in the US already takes 344 days.
That missing piece is adjudication, and it is the layer GenLayer exists to be.
We wrote the whole thinking behind it in The Compass, one of GenLayer's founding texts alongside the Manifesto and the whitepaper.
It lays out the worldview, the concept, and the mission in one place.
Read The Compass → portal.genlayer.foundation/g…






















