Independent Researcher #Web3. Building โ Educating โ Onboarding โ Improving RWA / stables / payments.
Global
Joined August 2022
- Tweets5.4K
- Following1.8K
- Followers1.8K
- Likes5.4K
Pinned Tweet
๐ ๐๐ญ๐๐ฉ-๐๐ฒ-๐๐ญ๐๐ฉ ๐๐ฎ๐ข๐๐ ๐ญ๐จ ๐๐๐๐ฅ-๐๐จ๐ซ๐ฅ๐ ๐๐ฌ๐ฌ๐๐ญ๐ฌ (๐๐๐๐ฌ): ๐
๐ซ๐จ๐ฆ ๐๐๐ซ๐จ ๐ญ๐จ ๐๐ง๐๐๐ซ๐ฌ๐ญ๐๐ง๐๐ข๐ง๐ ๐๐จ๐ค๐๐ง๐ข๐ณ๐๐ญ๐ข๐จ๐ง
Imagine owning a small fraction of a commercial building in New York, gaining exposure to US Treasury bills, or holding gold without personally storing a gold bar.
Now imagine being able to manage that investment through blockchain infrastructure.
This is the idea behind Real-World Asset (RWA) tokenization.
I manifesting this win IJN ๐
As much as I want to use every dollar to make my life better,
I couldn't forget the people who helped me get here.
So I took $1,000 and gave back to some of the people who supported me, believed in me, and played a part in making all of this possible
I couldn't reach everyone, but I wanted to show my appreciation in the little way I could
None of this would have happened without them, Adukes, and the Gamdom team
And by the way, I'll see you guys on my stream tomorrow
We've got a lot to celebrate and Gamdom Man first Max win
kick.com/thegamdomman
Stanley ๐ฝ retweeted
Removing liquidation does not remove leverage risk.
It changes the failure mode.
Conventional leverage can end suddenly at a liquidation threshold. The junior tranche from @2factorfinance has no conventional liquidation, but the position still has to overcome volatility drag and financing drag over time.
That is why the leverage ratio matters.
2Factor uses an assetโs long-run drift and volatility to estimate a productive band instead of applying the same 2x or 3x label everywhere. Bitcoin near 60% volatility points toward roughly 1.33x. Lower-volatility assets can support different levels.
The protocol is live on Base and working with Coinbase to bring this Kelly-optimized structure to tokenized stocks.
I joined the Points Program. Season 1 awards Marks for verified social activity, education and referrals. Purchases, deposits and holdings earn none. The top 10 split 1 BTC, paid in cbBTC:
points.2factor.finance/r/78cโฆ
Which risk is easier to evaluate: a visible liquidation threshold or drag that compounds over time?
In 2023 it was Peter Obis turn to support Atiku even if it meant to back out from the race
But you guys made it difficult for him, and even forced him to pull out from a political party with about 13 governors and several house members willing to support him
Today you're begging atiku to stepdown for him
On what ground?
Peter Obi is the most selfish politician in the history of Nigeria
That's why nobody is willing to give up anything for him
Atiku knows he doesnโt stand a chance of winning this election, but his desperation to stop Peter Obi from becoming president is pushing him to contest anyway.
Both major candidates from the 2023 presidential election are now united under the same banner, but Atiku has refused to align with them. He knows exactly what heโs doing.
What is he coming to promise again? Will he promise to rebuild Tudun Biri, a resettlement scheme that was already completed and commissioned in January 2026, months before his campaign flag-off? Will he promise to complete the Abuja-Kaduna road that his administration revoked, re-awarded at 752 billion naira, and now claims will be finished by November 2026, just in time for the election? Will he promise the Kaduna Light Rail, a project still in the mobilisation phase, with contractors yet to fully break ground? Will he promise the Mando-Birnin Gwari road, flagged off in August 2026 with a 24-month completion window, meaning it will not be finished until after the votes are counted? What exactly is new? What exactly is sincere? What exactly is not a campaign prop?
The Northwest is the de facto capital of poverty in Nigeria. The National Bureau of Statistics reports a multidimensional poverty rate of 75.8 percent in the region, second only to the Northeast. The Cadre Harmonisรฉ report shows 4.2 million people in the Northwest facing crisis-level food insecurity, with projections rising to 5 million in the 2026 lean season. UNICEF data shows primary school net attendance in the Northwest at just 47.3 percent, meaning more than half of school-age children are out of the classroom. Human Rights Watch, citing SBM Intelligence, recorded 2,938 kidnappings in the Northwest between July 2024 and June 2025, over 60 percent of the national total, with Kaduna alone recording 629 cases. Nigeria Watch recorded 3,974 deaths from rural banditry in 2025, more than double the figure for 2024. These are not political talking points. These are institutional figures. And yet, in the face of this avalanche of suffering, the president has been absent.
What will he tell the mother in Tudun Biri who lost her children to a Nigerian drone? What will he tell the father in Kachia whose family was abducted and never returned? What will he tell the farmer in Kajuru who buried his brothers after bandits attacked? What will he tell the flood victim in Zaria who lost everything? What will he tell the millions of out-of-school children roaming the streets of Kaduna while he commissions rail lines in Lagos? What will he tell the hungry, the poor, the abandoned, the forgotten? What will he promise again?
He will promise roads. He will promise rail. He will promise security. He will promise jobs. He will promise a better future. He will promise everything he promised in 2023 and failed to deliver. He will stand on a stage in Kaduna, surrounded by governors and senators and party chieftains, and he will tell the people that he cares. He will tell them that he has not forgotten them. He will tell them that the best is yet to come. And then he will fly back to Lagos, where the projects are real, where the money flows, where his people live.
Kaduna is not a campaign stop. Kaduna is a crime scene. And the man who is coming to campaign there is the same man who ignored the bodies, ignored the abductions, ignored the floods, ignored the hunger, ignored the children, ignored the pain. He ignored it all for three and a half years. Now he wants their votes. What will he promise again? What will he say that has not already been said? What will he do that has not already been abandoned? The people of Kaduna are waiting. But they are not waiting for promises. They are waiting for justice. They are waiting for a president who shows up when it matters, not when it is convenient. They are waiting for a leader who sees their suffering, not one who sees their votes. Will Tinubu be that leader? Or will he come, wave, promise, and disappear again?
Mohammed Bello Doka can be reached via [email protected]
Abuja Network News
Atiku should not have contested again
People like Aregbesola and Amechi are the ones giving him hope
These men can not influence more than 100 votes each
Stanley ๐ฝ retweeted
Two developers can solve the same problem without writing a single identical line of code.
We do not compare their source files character by character. We run tests and check whether both programs satisfy the required behavior.
AI needed a similar shift before it could work onchain.
Traditional blockchain consensus expects every node to reproduce the same output bit for bit. AI models can reach the same valid conclusion through different wording, so exact matching treats useful variation as disagreement.
With @GenLayerโs Equivalence Principle, each Intelligent Contract defines its own acceptance test. A leader proposes a result, and validators independently decide whether it satisfies the properties that matter.
The difficult part is no longer forcing AI to repeat itself. It is writing a test strong enough to accept valid variation without accepting a changed answer.
What would you test first: the final conclusion or the evidence used to reach it?
Two developers can solve the same problem without writing a single identical line of code.
We do not compare their source files character by character. We run tests and check whether both programs satisfy the required behavior.
AI needed a similar shift before it could work onchain.
Traditional blockchain consensus expects every node to reproduce the same output bit for bit. AI models can reach the same valid conclusion through different wording, so exact matching treats useful variation as disagreement.
With @GenLayerโs Equivalence Principle, each Intelligent Contract defines its own acceptance test. A leader proposes a result, and validators independently decide whether it satisfies the properties that matter.
The difficult part is no longer forcing AI to repeat itself. It is writing a test strong enough to accept valid variation without accepting a changed answer.
What would you test first: the final conclusion or the evidence used to reach it?
El-Rufai wanted to teach Tinubu a lesson in 2027 ๐ข
Who advised them to detain this gentleman ๐ฅฒ
Hey @grok how much has been invested in SW Nigeria by the Federal Government since 2023
And how much in total has been invested in the North entirely?
An AI agent that forgets your instructions is frustrating.
An agent that remembers the wrong instruction and quietly acts on it weeks later? That's a different problem.
Imagine a research agent tracking a tokenized Treasury fund. In August, it saves an old fee schedule. In October, the issuer changes the fees. The agent confidently retrieves its old note and publishes a fresh report using outdated figures.
Nothing crashed. The memory worked exactly as stored.
The failure was treating a retrieved memory as an unquestionable fact.
We're racing to give AI agents longer memories. But remembering more is only useful if an agent can recover the right information, know where it came from, and recognise when it needs correcting.
I've been studying that problem through @LighthouseWeb3's Memory architecture and independent research on long-term AI memory.
Here's what holds up, what needs qualification, and what builders should pay attention to. โ
There's an uncomfortable question behind every persistent-memory system:
What happens when the wrong information gets saved on purpose?
In AgentPoison (NeurIPS 2024), researchers demonstrated attacks that poison agent memory or retrieval knowledge bases, making compromised agents retrieve malicious examples and produce attacker-desired behaviour in the studied settings.
The lesson isn't that Lighthouse was compromised; the paper wasn't a Lighthouse evaluation.
It's that a memory system can faithfully preserve something an agent should never have trusted in the first place.
And forgetting deserves scrutiny too.
Lighthouse's documentation says "forget()" has different consequences depending on the backend. A batched record can disappear from the active index while its stored blob remains readable until the storage period expires. Memwal removes local state and unpins a mirror, but its encrypted blob may remain until expiry; the relayer has no direct deletion API.
That is not the same as universal, immediate data erasure.
If you're designing an agent handling financial, customer or confidential information, provenance, access controls, retention policy and deletion semantics belong in the architectureโnot in the small print.
Independent security research:
papers.nips.cc/paper_files/pโฆ
My test wouldn't be โCan it remember my favourite colour?โ
I'd give it an evolving project with decisions that change over time, then check six things:
1. Persistence: can it recover a saved decision after a fresh session?
2. Relevance: does it retrieve the right record, not just a vaguely similar one?
3. Freshness: can it distinguish the latest decision from an outdated version?
4. Provenance: can it tell me where the record came from?
5. Integrity: can stored content be checked for unexpected alteration?
6. Control: what really happens when the memory is forgotten?
Those are proposed evaluation criteria, not claims that I've completed a Lighthouse benchmark. They also align with the concerns raised in NIST's AI Risk Management Framework: reliability, transparency and privacy are system-level responsibilities.
NIST: nist.gov/itl/ai-risk-managemโฆ
To me, Lighthouse's most interesting contribution is separating memory from the model and making it recoverable, portable to authorised clients and content-addressable. Its own documentation also exposes the trade-offs we should be asking about.
The next step for agent memory isn't โremember everything.โ
It's remember what matters, retrieve it when needed, verify the record, update the factโand know when to forget.
Explore the implementation: memory.lighthouse.storage/
Technical docs: docs.lighthouse.storage/memoโฆ
What's the bigger risk for autonomous agents: losing context or trusting stale context?
โ @StanleyCrypt_
