Nelson Ijih

Clear eyes Full hearts Can't lose

[Nelson Ijih @NelsonIjih]

Failure humbles you if you were doing it for the acolades.

[Nelson Ijih @NelsonIjih]

Whenever they ask "how many more reps you have to go" at the gym so they can take over the machine

My response is usually.." About 10 mins"

[Nelson Ijih @NelsonIjih]

Checkout @RobinhoodApp $HOOD onchain stock tokens activities.

Current snapshot as of today:

- 1,998 total stocks tokens deployed
- 590 active (i.e onchain supply > 0)
- $10m+ onchain marketcap.

https://t.co/5sfdcrwTsW

[Nelson Ijih @NelsonIjih]

- WASM bytecode execution engine for high-perf specialized L2 seqeuencer for WASM VM blockchains so smart contracts execute at rate of network card packet ingestion.

- High-perf transactions conflict detection for SVM parallel transaction scheduling.

[Nelson Ijih @NelsonIjih]

Building apps for FPGA has been one of the impediments to it's adoption.

multiple paradigms have been tried with limited success(e.g openCL, VLS,etc).

With LLMs, RTL code generation for FPGAs will likely see an uptick.

some use case includes:

[Sally Ward-Foxton @sallywf]

As part of a refocus on core parts of the business, @intel is to sell a majority stake in FPGA maker @AlteraFPGA_ to @silverlake_news for half what it paid. Some analysis drawing on @eetimes reporting from the last decade:
https://t.co/m8BY1bF4Rx

[Nelson Ijih @NelsonIjih]

benchmark from this paper

https://t.co/8Zx6joo9X0

[Nelson Ijih @NelsonIjih]

Over head of TEE-enabled LLM queries are negligible, looks like we'll soon be seeing "Inside a TEE" lock icon (similar to https in browser) in LLM chat windows?

[Nelson Ijih @NelsonIjih]

The year the #TEE gets hotter.

[Nelson Ijih @NelsonIjih]

@eigen_da vs other DAs

[The Figen @TheFigen_]

Daughters of the wind!

https://t.co/GTFigd6cUR

[Nelson Ijih @NelsonIjih]

@nikzh 😂

[Nelson Ijih @NelsonIjih]

Copyrightable = Proof of human work.

Judge rules U.S copyright “protects only works of human creation” which is "at the core of copyrightability, even as that human creativity is channeled through new tools..."

https://t.co/B00K4NzCE7

[Nelson Ijih @NelsonIjih]

A few weeks(even days) of fully onchain gaming with millions of players would generate chain size more than the nearly a decade old ETH chain size of 1TB

Appchain use case

[Messari by Blockworks @MessariCrypto]

Onchain games represent a completely new architectural shift in gaming.

Also known as autonomous worlds, onchain games have their game state and logic fully onchain, allowing developers and users to take full advantage of the open and composable nature of blockchain infrastructure.

Where traditional game development is an extremely costly and time-consuming endeavor, onchain games unbundle and open up the gaming stack. Each component is then able to be reused or modified by developers.

The modular approach taken by onchain games allows for reduced development costs and quicker experimentation. Additionally, the open architecture enables anyone to build games, items, or characters on top of a shared world layer.

The infrastructure layer for autonomous worlds is currently led by:

~ @latticexyz's MUD engine
~ Dojo
~ @0xcurio's Keystone OPchain
~ @ArgusLabs_’ World Engine.

The global gaming market is massive, valued at $250 billion in 2022, representing a large and growing market for new onchain games to exist in.

[Nelson Ijih @NelsonIjih]

Though full user privacy can be achieved with this approach, however, the user may "skip" the missing part if they can live with less accurate results to avoid interacting with FHE-enabled server side.

[Nelson Ijih @NelsonIjih]

It'll take ~1.5 hours to process a user query if chatGPT proprietary model were to be protected & a user's query be made completely private using FHE(fully homomorphic encryption) according to recent numbers by @Ingo_zk

I wrote about some killer applications of FHE in 2020:

[Nelson Ijih @NelsonIjih]

Some of the killer applications for #FHE (fully homomorphic encryption)are:

#Blockchain "send" transactions by wallet providers to decrypt encrypted private keys only in #FHE enclave to "send" funds

#IoT public edge devices that host & process proprietary/sensitive AI models

[Nelson Ijih @NelsonIjih]

@zama recently suggested partial model protection by deploying parts of the proprietary model(unprotected )on user's device to partially process the user's query locally and the rest of it in the cloud in FHE mode.

https://t.co/zRYTt2iGm8

[Nelson Ijih @NelsonIjih]

SGX bug seem to affect only three CPUIDs?

https://t.co/aIpDNGbREu

[Nelson Ijih @NelsonIjih]

Being horrible at apparel styling, I wish there's some #Dalle + Google Bard + #GoogleLens to help me step up my fashion styling game with shoppable links🤓

[Nelson Ijih @NelsonIjih]

Especially given that shuttle programs at foundries like TSMC getting, not looking at ASIC option at this scale would be foolish...Not many GPUs to go around.

[martin_casado @martin_casado]

If a company is going to train a $5B model, wouldn't it make sense to use 5% of that to build a custom ASIC for that model? The benefits are certain to dwarf the 5% investment.

At these scales, even ASIC design costs become marginal.

[Nelson Ijih @NelsonIjih]

TradFi'ing DeFi to some extent is good balance for path forward

https://t.co/NHY26y8CEE