Dev Dahniel

Mad scientists @sandwormlabs

[Dev Dahniel @dev_dahniel]

@0xKofi @0xBoxer @andrewhong5297 Hey kofi, please check your DMs when you get a chance 🙏

[mert @mert]

leaving crypto for ai is incoherent

what are you going to do? train a frontier model? no

call an API provided by the big labs? yes

groundbreaking stuff

[who decided that? @destinyspoke]

Long live Nigeria🕺🏽

[mert @mert]

> lost mom at young age
> started working at 8yo
> was on welfare, slept at random churches for shelter, getting by with cream cheese packets kek
> went to worst ranked school in the state
> huge student debt 6y ago
> now, bought dad a house!

just keep going

[Dev Dahniel @dev_dahniel]

@TFAwinner @unicodeveloper I totally agree

[Dev Dahniel @dev_dahniel]

@seunlanlege And they act like post-quantum cryptography doesn't exist

[Dev Dahniel @dev_dahniel]

@seunlanlege Crypto bros like to exaggerate the tech and they tend to be doomers.

[Web3 Philosopher @seunlanlege]

It's so funny that actual physicists say quantum computers are overhyped, while crypto bros run around screaming "post-quantum" on top of their lungs.

[who decided that? @destinyspoke]

With regards to this

@dev_dahniel should have an article coming out soon

On why what we need and what businesses require the most isn’t AI replacement, but AI automation with human supervision

We saw this while working on @sandworm

AI right now, LLMs follow the path of “plausibility” and has not fully grasped the concept of

“this is not strategically important”
“These numbers are not correct”

Without guardrails

It almost never says “I don’t know” it will rather find a way to give you something, even if that something is wrong under scrutiny

It will happily hallucinate numbers and information for you

And in enterprise where workflows can depend on other workflows, this is a disaster if the false output of one agent is used to feed the input of another agent

He spent time detailing to me how the directions that many companies went with AI was bad, a competitor we looked at had no guardrails and as a result their outputs where very basic and prone to hallucinations

Here is a preferred workflow

- AI works in chunks(connect windows are getting better, but for now, chunks work best)

- output becomes “accept” “deny” or “refine”

So a human in the loop system

Sounds dystopian, but what can a man do? Your job is essentially to course correct and present errors that can wipe billions in valuation.

[EnochMbaebie.eth | ETHLagos| LBW 2026 @Enoch_Mbaebie]

So thrilled to announce our community's official acceptance into the @ETHGlobal
family!
Feels great to be among 65,000+ new community members from 153 countries
Another win for https://t.co/FXSref6GUn,
let's 10X our efforts and keep building the decentralized future together!

[Dev Dahniel @dev_dahniel]

In summary, we’re creating a library of community-curated blocks and hard-coded primitives that you can assemble to perform composable analysis. What has been built out is a notebook specifically for on-chain analysis

[Sandworm @sandwormlabs]

On-Chain Analytics as LEGO Bricks https://t.co/4XE9Fmwd4r

[Dev Dahniel @dev_dahniel]

We've spent five months building Compostable Analysis, and I have to give it up for @destinyspoke. He has articulated it far better than I ever did.

[who decided that? @destinyspoke]

1.5million onchain queries exposed insight that we never knew existed

@dev_dahniel is feeling modest, so I'll do the write up

Over the last couple of months we've been building @sandwormlabs , what started out as a simple AI powered analytics platform turned into very deep insight into onchain data, from finding out what most people searched the most, to how people behave onchain and more

Looking at data daily will force an understanding into you, an understanding that most queries and info people seek onchain(even the top data analysts) are the same 90% of the time. So we asked the question, why dont we reduce redundancies by building the info like lego bricks, if what we are doing works, we would be essentially proposing a new standard model for Onchain analysis
Follow @dev_dahniel
Follow our github https://t.co/CDA8KYcNBy

cc @Sir_Damilare @jessepollak @codingcas @dicethedev

https://t.co/7X9NPezUe3

[who decided that? @destinyspoke]

1.5million onchain queries exposed insight that we never knew existed

@dev_dahniel is feeling modest, so I'll do the write up

Over the last couple of months we've been building @sandwormlabs , what started out as a simple AI powered analytics platform turned into very deep insight into onchain data, from finding out what most people searched the most, to how people behave onchain and more

Looking at data daily will force an understanding into you, an understanding that most queries and info people seek onchain(even the top data analysts) are the same 90% of the time. So we asked the question, why dont we reduce redundancies by building the info like lego bricks, if what we are doing works, we would be essentially proposing a new standard model for Onchain analysis
Follow @dev_dahniel
Follow our github https://t.co/CDA8KYcNBy

cc @Sir_Damilare @jessepollak @codingcas @dicethedev

https://t.co/7X9NPezUe3

[Sandworm @sandwormlabs]

On-Chain Analytics as LEGO Bricks https://t.co/4XE9Fmwd4r

[shafu @shafu0x]

I actually thought I would never say this but combining AI and crypto actually makes fucking sense now

[Dev Dahniel @dev_dahniel]

@destinyspoke ❤️

[Dev Dahniel @dev_dahniel]

@a_ix_d @el_humorrz @HarriObi 🤣

[Dev Dahniel @dev_dahniel]

@dannithomx It's an amazing night, with different stories unfolding on various fronts. There is a common denominator. Having an AI gf is a strategy to avoid being in any of these situations

[Dev Dahniel @dev_dahniel]

Remember, your AI girlfriend won't take your conversations to Twitter. Stay happy and enjoy having an AI girlfriend.

[Dev Dahniel @dev_dahniel]

@lifeofdanel This is some Death Note type shit, like I am just entertained