AI-meda Research

Group of AI Researchers and Crypto Natives Focusing on Decentralized AI Research

[AI-meda Research @AIMedaResearch]

D/AI Projects Upcoming Launching (Q3 ~ Early Q4)

🟥 Incentivized Testnet / Mining / Point Program:
@OraProtocol, @Lab7007, @heurist_ai, @0G_labs, @sophon, @getgrass_io, @AlloraNetwork, @AIWayfinder, @nubit_org, @Lilypad_Tech, @Rivalz_AI, @ShinkaiProtocol, @kuzco_xyz, @CreatorBid

🟩 TGE / Project Launched:
@Lab7007 (expected), @nubit_org (expected), @Lilypad_Tech, @AethirCloud, @myshell_ai, @Rivalz_AI, @ChasmNetwork, @Privasea_ai

🟨 Testnet Launched:
@gensynai (expected), @sentient_agi (expected), @AIWayfinder, @nesaorg, @NeuroMeshInfo, @flock_io, @Neura_io

🟦 Token, Node Sale:
@nillion, @Privasea_ai

* Referred to the structure of @layerggofficial's posting

[Tony Lim @Tonynsight]

Some updates in deAI(with @AIMedaResearch)

1. AI eco databases
https://t.co/HvINm5TbbV

2. AI x Crypto communities by @jasonzhao
https://t.co/6PT4ajAcZk

3. Anyone participate in @Privasea_ai node sale?
https://t.co/K0hu5q9UpK
https://t.co/ia3YUMns9I

4. @OraProtocol received retro fund by optimism
https://t.co/yFXEBGYf5n

5. @PhalaNetwork in @MessariCrypto report, hackathon list
https://t.co/UYUyxSefge

https://t.co/qhV0GlhbuT

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #7
Heurist - Community owned AI DePIN

About Heurist
@heurist_ai is a Layer 2 network for AI model hosting and inference, built upon ZK Stack

- Heurist provides serverless access to open-source AI models hosted by a decentralized network of computing resources; The HuggingFace of Web3

- Heurist empowering crypto enthusiasts to engage with AI, and enabling AI model developers to access more compute resources, publicize their models, and earn revenue share

Workflow
Heurist participants include Model Users (Consumers), Miners (Model Hosts), Model Creators, Application Integrators, and Validators

1. Model Users pay a fee to request inferences using AI models which Model Creators made, such as text generation, image generation

2. Miners utilize their own GPU resources to run models and handle requests from Model Users

3. Validators periodically check the correctness of the data produced by miners; If a miner is found to be producing invalid or fake data, their stake is slashed

Data Privacy
Inference using sensitive information such as health records, identity data, etc.. input data must be encrypted

- Heurist addresses this issue by categorizing Miners into Public Miners and Privacy-enabled Miners

- Privacy-enabled Miners handle encrypted requests, while Public Miners process standard requests

Privacy-enabled miners are trusted third parties whose trust can be established through two primary methods

- Off-chain consensus, such as IRL laws or agreements, which is technically straightforward to implement

- Utilizing TEE to ensure secure and confidential processing of sensitive data (it is expected that TEE will support AI workloads in the future)

Tokenomics
$HUE is a utility token with a dynamic supply, influenced by both emission and burning mechanisms

- $HEU max supply: 1,000,000,000 (1b)

- Emission Mechanisms: Mining, Staking

- Burning Mechanisms: When users pay for AI inference, a portion of the HUE payment is permanently removed from circulation

Incentivized Testnet
Heurist allocated 5% of total $HUE token supply for incentivized testnet mining rewards (season 1)

- You can earn Llama points by running LLM models and Waifu points by running Stable Diffusion models

- 2.5% of the tokens will be linearly distributed for each type of point

- Season 1 will last until the end of July

- Miner Guide: https://t.co/1kKUBBD8O2

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #6
Nillion - The First Blind Computer

About Nillion
@nillion is a secure computation network that decentralizes trust for high value data in the same way that blockchains decentralized transactions

- Nillion provides a a secure network for AI data privacy

- Nillion integrates and orchestrates Privacy-Enhancing Technologies (PETs)

- Nillion executes computations on data without decryption (Blind Computation)

The Orchestration Layer
- Maximizes the strengths of different privacy technologies (ZKP, FHE, MPC) by integrating them, assuming no single encryption can solve all problems

- Provides developers with access to a combination of PETs (Privacy-Enhancing Technologies) like Nil Message Compute (NMC)

Blind Computation
- Allows computations on encrypted data without decryption

- Keeps sensitive data encrypted during executing, preventing exposure

Nillion x Coinlist
- Nillion will launch a community round on Coinlist on June 19th

- Community round allows public investment in the same terms as the angel and VC rounds

- Everyone who (properly) completes The Nill Pill before that date will be given priority access

Fundraising
- Secured $20M in seed funding on December 12, 2022

- Investors: Distributed Global(lead), Big Brain Holdings, Chapter One, Hashkey, Truffle Ventures, OP Crypto
(full investors: https://t.co/hPgzqEsdbR)

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #5
Wayfinder - An AI Agent for Web3

About Wayfinder
@AIWayfinder provides a Shell, a user-owned LLM model, enabling onchain operations based on user’s intent entered in natural language

- Users can use a Shell to operate complex onchain operations such as crosschain swaps and purchasing NFTs of specific rarities, by paying $PROMPT

- A Shell can process user intents using the Path, stored in the Wayfinder network; shells pay $PROMPT to the path creator for using the path

Web3 AI Agent
Wayfinder provides a Shell that allows users to engage in web3 activities in the form of a chatbot

- Wayfinder overcomes the limitation of existing AI not being able to engage in Web3 activities to create value
(e.g. Swap 80% of my SOL from Solana for ETH from Base with a best efficiency)

- Wayfinder can make meaningful contributions to the development of cross-chain interoperability solutions that significantly reduce the complexity of blockchain environments

- Users can specialize their shell by learning specific knowledge or personalize it by giving it personality, and can also sell the memory of the learned shell

Wayfinder Network
Wayfinder Network graphs the Web3 ecosystem, allowing shells to explore paths

- Protocols, contracts, token standards, etc. (e.g. Uniswap, Aave) become nodes, and the functions and routines for calling each (e.g. swap, approve) become edges

- If shells use a path that has already been created, a token reward is given to the path creator, and if there is no path, a new path can be created by exploring the graph

- Recently, on June 1st, first lending routes using Aave were created (also Improved agent transaction reliability, composability)
(tweet: https://t.co/Vv5yGX9fpy)

Tokenomics
- Total Supply: 1b (expected)

- Monthly distributed to ecosystem token holders ($PRIME) who cache their tokens within Wayfinder across 3 years

- Potential $PROMPT Genesis Allocation
Cached: 40% / Future Incentives: 5% / Wayfinding Rewards: 5% / Foundation: 6.66% / Launch Partner 1.34% / Team: 16.51% / Investors: 25.49%

Fundraising
- Total Raised for @ParallelTCG (Which launched Wayfinder): $85M

- Raised $50M with $500M valuation in the funding round, by Paradigm (Oct 22, 2021)

- Raised $35M in the funding round, including Solana Ventures, Base, VanEck, ... (Mar 28, 2024)

[Accelxr 👾 @accelxr]

Crypto x AI Weekly #3
​
Covering the recent developments in AI x Crypto, including news, interesting reads, upcoming events, and more.
​
Dive in below 👇

💾 Compute 💾
💾 @bittensor_ Dynamic $TAO Demo is Here {https://t.co/QoJzMFD70Q}
💾 @khoomeik dives into "covered model" definitions in SB 1047 {https://t.co/YFxp0TR64p}
💾 Introducing the @Livepeer AI Subnet {https://t.co/BwJZoFrpO1}
💾 Image gen model fine-tuned on @akashnet GPUs by @itsandrewgao {https://t.co/qVoooOMqVT}
💾 'ChaAI' by @sigmoidwtf: Empowering Chain Abstraction for AI Networks {https://t.co/yn0NPjoQD4}
💾 Elon Musk to use 100,000 H100s for Grok {https://t.co/4JyUbZD6gf}

🤖 Agents 🤖
🤖 @virtuals_io releases protocol update {https://t.co/CpN13YWV1e}
��� @MorpheusAIs Community Website {https://t.co/cI2byEz8Xj}
🤖 @BrianknowsAI releases ENS actions {https://t.co/rkd9DvaLst}
🤖 @autonolas launches Pearl - a staking and agent launching application {https://t.co/zhkC4lbBhT}
🤖 @Spectral_Labs showcases Onchain Explorer Agent {https://t.co/Zj2oZL6qc4}
🤖 @polywrap_io launches a Farcaster based agent called Indexer {https://t.co/NqLrpCaG9a}

🔐 Verification & Privacy 🔐
🔐 Predicting liquidation risk with @gizatechxyz {https://t.co/93I2jl2Zzh}

🌐 Other Infra 🌐
🌐Near establishes @near_ai, an R&D lab led by @ilblackdragon & @AlexSkidanov {https://t.co/mO3srbBLRW}

🎮 Gaming 🎮
🎮 Microsoft adding AI to Minecraft and other titles {https://t.co/N1HrvRV8UX}
🎮 @todaythegame holders to receive @AethirCloud airdrop {https://t.co/zeIm6EvVdz}

🎨 Art 🎨
🎨 @imgn_ai launches new Funime anime/manga model {https://t.co/Cas91bzWlr}
🎨 'p5.botto WIPs' by @bottoproject displayed by @verse_works {https://t.co/AuU8OCmIKL}
🎨 @rainisto public sale of 'Vacation' on @verse_works begins {https://t.co/GMJv2MYciK}
🎨 'AI Index' by @obv_ious & @ivonatau minting now on @verse_works {https://t.co/rQeXS1Z7oG}

🚜 Points, Incentives, & Airdrops 🚜
🚜 @GenLayer points program on Galxe {https://t.co/slF8YBYoec}
🚜 @AethirCloud Drop ends today {https://t.co/9sRyX9idnv}

🧠 AI Research 🧠
🧠 Foundational model wars over the past 12 months {https://t.co/qcrAHNUk7P}
🧠 MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning {https://t.co/7WgHpE2zJP}
🧠 OpenRLHF: An Easy-to-use, Scalable and High- performance RLHF Framework {https://t.co/Srx7kM1b6O}
🧠 FIFO-Diffusion: Generating Infinite Videos from Text without Training {https://t.co/hZ9uQXdkmi}
🧠 Your Transformer is Secretly Linear {https://t.co/fwQbvkY9PX}
🧠 Agent Planning with World Knowledge Model {https://t.co/9vZaN7o3U9}
🧠 Risks and Opportunities of Open-Source Generative AI {https://t.co/B7YsJfgNRG}
🧠 Extracting Interpretable Features from Claude 3 Sonnet {https://t.co/EIVleeIOJ5}

🖥️ Interesting Reads & Listens 🖥️
🖥️ @DWFVentures releases crypto x AI ecosystem map {https://t.co/fRBwC2OUNu}
🖥️ Democratizing AI with @gensynai's Decentralized Compute {https://t.co/6F0WtBPqZF}
🖥️ @NoaNuman provides a breakdown of @AIWayfinder {https://t.co/UfWr46fR6p}

🎟️ Upcoming Events / Hackathons 🎟️
🎟️ AI Film Awards hosted by @aiFilmAcademy at @NFCsummit, May 28th (https://t.co/s05fZF2C3m)
🎟️ Freethink & Drink: Decentralized AI Happy Hour With Wire Network, Morpheus & 6079, May 29th {https://t.co/g4ToiWWotf}
🎟️ Austin on my MIND with @0G_labs, @AethirCloud, @AlloraNetwork, @flock_io, @iotex_io by @mindnetwork_xyz {https://t.co/F8E8gA9Vby}
🎟️ DeAI Day Austin by @MorpheusAIs, @HelloLumerin, @TryVenice & @spns_ai - May 30th (https://t.co/MpGDmf77sg)
🎟️ AI Summit at Consensus, May 30th {https://t.co/D1glRIr2Qx}
🎟️ The Future of Compute with @QuaiNetwork, @ionet, & @HelloLumerin, May 30th {https://t.co/ezqW3dQCBJ}

💊 Hallucinations 💊
💊 Infinitely Recursive of Game of Life {https://t.co/73fLkWt7vu}
💊 CNN for digit prediction (MNIST) in Minecraft {https://t.co/XAzymovv0h}
​
🔥 Last week's update icymi: https://t.co/9IxM3kDAoo
​🔥 Also, I will be publishing an overview of onchain AI Agents this week so be on the lookout 👀

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #4
https://t.co/hVG9zVyDoa - The Internet of GPUs

About https://t.co/hVG9zVyDoa
@ionet is a GPU DePIN project that can provide substantial computing resources by clustering GPUs

- https://t.co/hVG9zVyDoa can provide GPU resources at a much lower cost compared to centralized cloud computing platforms like AWS or Azure

- You can provide surplus GPU resources and receive $IO tokens as a reward. If you need GPU resources, you can rent them with just one click

GPU Clustring
- GPUs are clustered by geographic location, and mesh VPN is used to minimize latency

- IO Tunnels (Reverse Tunnels) are used to secure data privacy in connections, simplify the process, and make it easier for https://t.co/hVG9zVyDoa to manage each IO Worker

- Using the web3 token reward system, it can provide computing resources at a lower cost than centralized platforms

Running a node
You can operate an IO Worker by providing GPU

- Season 2 is underway; you can earn IO points by operating an IO Worker, but season almost done (~ 05.31.)

- You can provide various GPUs/CPUs including NVIDIA GeForce Series, Apple M1/M2/M3, and AMD Ryzen Series

- Guide link: https://t.co/mWQot3Okpc

Tokenomics
- Total Supply: 800m / Initial Supply: 500m

- $IO uses a programmatic coin burn system

- After all $IO distributed (After 20 Years)
Community: 50% / Early Backers (Seed): 12.5% / Early Backers (Series A): 10.2% / Initial Core Contributors: 11.3% / R & D Ecosystem: 16.0%

Fundraising
- Series A Raised: $30M / Valuation: $1B

- Raised $30M in the Series A, led by Hack VC (Mar 05, 2024)

- Full investment info: https://t.co/6Ou0aU1axt

Reference: AI-meda Research, https://t.co/hVG9zVyDoa

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #3
Grass- The Data Layer of AI

About Grass
@getgrass_io is a DePIN project for data, focusing on the collection, storage, preprocessing, and sale of data

- Grass is currently focusing on the decentralization of data collection; Decentralization in other areas is also planned for the future

- Grass is a Solana Layer 2 that posts proofs(zkp) of their works on Solana

Importance of data
- While training an AI model is obviously important, the answers you get when you use it are based solely on the correlations it finds in the training data

- Quality data collection is becoming difficult, Grass’ DePIN enables data collection at an affordable price

Workflow
1. Client requests data crawling for the Target Web Server

2. Validator makes TLS handshake with Target Web Server and requests Grass Nodes to relay encrypted packets

3. Data is decrypted with the Validator's session key and delivered to the Client

4. Data is stored in the Grass Data Ledger and settled as ZKP goes up to Solana

Running a node
We can operate the Grass Node for free and get Grass Points as a reward; Epoch 5 is underway

- Download link: https://t.co/xZMUgFoyNI

- Currently only data crawling is possible, but data labeling will also be added

Fundraising - Total Raised: $4.5M
- Raised $1M in the pre-seed round, led by No Limit Holdings (Jul 04, 2023)

- Raised $3.5M in the seed round, led by Polychain Capital and Tribe Capital (Dec 18, 2023)

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #2
Privasea - AI Network using FHEML

About Privasea
@Privasea_ai is an AI + DePIN Network enabled through FHEML (Fully Homomorphic Encryption + ML)

- Privasea supports various functions related to FHE ML, including AI inference, model training, etc..

- Privasea AI Network tries to adhere to data protection regulations, including the GDPR (General Data Protection Regulation)

FHE / FHEML
- FHE is a cryptographic scheme where computations are preserved

- ex) a + b = a + b <-> enc(a) + enc(b) = enc(a+b)

- FHEML is a method that performs operations used in ML on encrypted data using an FHE scheme

Workflow
1. User encrypt the input data with FHEML library

2. The designated Privanetix node receives encrypted data, performs ML and makes zkp (zkML currently not implemented / only generate results)

3. The results are decrypted by the decryptor and delivered to the user (receiver)

Tokenomics
- Distribution
Mining, Staking: 45% / Team: 10% / Backer (VC, Partnership): 20% / Marketing, Community: 15% / Reserve: 6% / LP: 4%

- Token utility
1. Using Privasea AI service
2. Rewards for Privanetix nodes, decryptor provider, etc..
3. Staking for network security, governance, etc..

Fundraising
- Raised $5M in the seed round, including Binance Labs (Mar 04 2024)

- Strategic investment from OKX Ventures, Laser Digital, Tané (Apr 03 2024)

You can check out our group's detailed research on Privasea at the following link: https://t.co/MGJYO1CeBS

Reference: AI-meda Research, Privasea

[AI-meda Research @AIMedaResearch]

De/AI project Spotlight #1
ORA - Ethereum's trustless AI

About ORA
@OraProtocol is Ethereum’s trustless AI that enables AI and any-size compute on blockchains

- Ora ensures verifiable ML by verifying and proving that the requested inferences have been performed by the respective models

- Ora enables sustainable funding for open-source AI models and grants users ownership through IMO: Initial Model Offering

Verifiable ML
Using opML, zkML and opp/ai, ORA ensures that the inferences have been performed by the respective model (Similar to optimistic / ZK rollups method)

IMO (Initial Model Offering)
IMO tokenizes AI model onchain. Token holders will receive the benefits of revenue streams:
Model Ownership (ERC-7641) + Inference Asset (ERC-7707)

Fundrasing
- Ora raised $3m in Pre-Seed round, led by dao5 and Hongshan(Sequoia China)

You can check out our group's detailed research on ORA at the following link:
https://t.co/KRumjatFJL

Reference: AI-meda Research, ORA, Decipher

[AI-meda Research @AIMedaResearch]

D/AI Projects Industry Map

1. Middleware Layer Application
🟨Featured on Binance Research:
FetchAI, SinfularityNET, Ocean, Worldcoin, Orachain, Modulus

🟩Other notable projects:
Chasm, Heurist, Ritual, Ora, Phala, Privasea, Kuzco, Allora, Lilypad, KIP Protocol

Projects that utilize the resources from infrastructure projects to provide products and services like AI agents, inference verification, authentication, etc.

2. User-facing Application
🟨Featured on Binance Research:
Arkham, Numerai, BitsCrunch, Mozaic, PAAL, SleeplessAI, 0x0

🟩Other notable projects:
Morpheus, Myshell, Autonolas, CreatorBid, Wayfinder

Projects that leverage the power of AI to offer consumer-facing use cases such as trading, image generation, data analytics, gaming, etc.

3. General purpose compute
🟨Featured on Binance Research:
Akash, Render, Nosana, Cudos, PAAL, 4EVERLAND, Flux, Latitude, Bacalhau, Golem, DeepBrainChain

🟩Other notable projects:
Ionet, Aethir, Grass, Rivalz

Projects that provide (GPU based) computing resources for model inference or data scrapping based on a decentralized infrastructure and crypto-incentive mechanism.

4. Model Training/Fine Tuning
🟨Featured on Binance Research:
Bittensor, gensyn, together AI, enqAI, HHypercycle, Nimble

Projects that focus on training computer systems to make inferences based on data and algorithms.

Reference: AI-meda Research, Binance Research(https://t.co/d4267jPe83)

[AI-meda Research @AIMedaResearch]

D/AI Projects Upcoming Launching (Q2 ~ Early Q3)

🟥 Incentivized Testnet / Mining / Point Program:
Heurist, Ora, Grass, Kuzco, GM Network, Morpheus, Allora, Lilypad

🟩 TGE / Project Launched:
Heurist(expected), Aethir, Morpheus, Chasm, https://t.co/7K7fPSevqk, https://t.co/Dp32uQh2sa, Inferix, Nimble

🟨 Testnet Launched / Ongoing:
Shinkai, Neura, Prime Intellect, Nimble, Allora, Lilypad, Privasea AI, 0G Labs, Myshell

🟦 Node Sale:
Rivalz, Sophon

⬛️ Lack of info but paying attention:
GatlingX, Gensyn, Sentient

[AI-meda Research @AIMedaResearch]

2024.04.17~2024.04.19 daily summary for public - 국문

🗣 Main Discussion

Morpheus
-Morpheus 페어 런치 플랫폼: https://t.co/ZflS99abOw

0G
-여러 샷아웃들을 받는중
-tweet link https://t.co/9RFlkYEjFW

🖥 Node Running/Mining

Heurist
-노드 수가 1200개에서 밤 사이 3000개로 증가함
-링크에서 전체 노드수를 확인할 수 있음: https://t.co/wa8NageFuX

[gonedark @goinggddark]

Study crypto x ai conferences, study @mheinrich. One of the most active founders rn, giving every discussion a go + on top of that really enjoyable listening to him.

Not surprised he‘s involved in basically anything ai currently, doesnt matter if advisor or ecosystem-wise.
0G on🔝

[AI-meda Research @AIMedaResearch]

2024.04.17~2024.04.19 daily summary for public - English

🗣 Main Discussion

Morpheus
-Morpheus fair launch platform: https://t.co/ZflS99abOw

0G
-Shout out by @goinggddark
-tweet link: https://t.co/9RFlkYEjFW

🖥 Node Running/Mining

Heurist
-node increased 1200 to 3000 overnight
-can check total node amount here: https://t.co/wa8NageFuX

[gonedark @goinggddark]

Study crypto x ai conferences, study @mheinrich. One of the most active founders rn, giving every discussion a go + on top of that really enjoyable listening to him.

Not surprised he‘s involved in basically anything ai currently, doesnt matter if advisor or ecosystem-wise.
0G on🔝

[ForumAI @ForumAILabs]

🚀 Exciting news! ForumAI has secured a $200k funding round, backed by @ark_stream @County_Cap @AIMedaResearch and @Normie_Capital This investment boosts our journey in decentralized AI, focusing on product development, marketing & community building.

Blog: https://t.co/7LO3csxoVq

[AI-meda Research @AIMedaResearch]

* Heurist 아티클 링크가 변경되었습니다

국문: https://t.co/KCPMpNiV9x

영문: https://t.co/uv2eImxztF

[AI-meda Research @AIMedaResearch]

2024.04.13~2024.04.14 daily summary for public - 국문

🗣 Main Discussion

GPT-5, AI agent 시장에 대한 생각들과 보안 문제
-GPT-5의 등장으로 AI agent 시장은 엄청난 성장을 경험할 것으로 예상됨
1. Future AI technology에 대한 두 가지 bottleneck:
컴퓨터 기술적인 문제 (e.g., FHEML, ZKML, 높은 성능의 LLMs)
2. 보안 문제
보안 문제
-사용자 권한(이메일 접근, 운영체제 등에 대한 접근 등)에 대한 잠재적 위험과 보안 문제가 우려됨
-그의 경험에 따르면 보안 문제는 AI 발전에 매우 중요함
AI Agent
-GPT-5의 등장으로 Web3 AI 에이전트 프로젝트들이 상당한 성공을 거둘 것으로 기대함
-하지만, AI Auditing DAO, H2E (Hacking to Earn)등에 대한 비즈니스 모델도, 높은 포텐셜과 함께 등장할 수 있음

Bittensor
-Bittensor의 S14는 LLM들에 대한 공격을 방지하는 것들을 하고 있고, 위의 보안과 연관되어 보면 좋아보임

📚 AI Project Research

Heurist에 대한 아티클 by Junho Kim, L Tony - Full article link: https://t.co/29RZzrvt5O

Short Summary
-Heurist를 오픈소스 AI Model 카테고리로 분류할 수 있고, Web3 dApp들에 모델을 제공해줄 수 있는 허브와 같은 역할로 기대함
-수백 개의 M/B 파라미터를 보유한 독점 모델과 오픈 소스 모델을 비교하기는 어려움
-오픈 소스 모델은 적은 수의 매개변수를 사용하여 맞춤형 모델을 개발하려는 사람들에게 수요가 있을 것으로 예상됨
-Crypyto-based Model들을 Heurist에서 배포, 통합하고, AI tool을 제공할 수 있다면, Web3에서의 Hugginface가 될 수 있을 것으로 기대됨
-Heurist는 AI Model 호스팅 및 추론을 위해 ZK 스택으로 ���발된 L2 네트워크
-구성요소로 Consumers, Miners/Model Hosts, Model Creators, Application Integrator, Validators 가 있음
-2024년 3분기에 토큰이 나올 것으로 기대됨
-비슷한 컨셉의 경쟁 프로젝트들: Ritual, Neura, Nimble, Allora, Shinkai

Heurist에 대한 부가적인 bullish points들
-몇 안되는 Hyperstack(ZK stack으로 개발된) 프로젝트
-zksync팀이 푸쉬해줄 수있고, zksync tge에 간접수혜 가능성
-명확한 tokenomics & token distribution → 좋은 퍼포먼스를 보일 수 있음
-몇달간 커뮤니티를 보면서, 팀이 열심히 일하는 것이 느껴짐
-프로덕트도 꾸준히 나오는 중
-펀딩라운드가 괜찮은 밸류로 마무리되면, 상당히 괜찮을 것

[AI-meda Research @AIMedaResearch]

* Heurist article link modified

Eng: https://t.co/uv2eImxztF

Kor: https://t.co/KCPMpNiV9x

[AI-meda Research @AIMedaResearch]

2024.04.13~2024.04.14 daily summary for public - English

🗣 Main Discussion

Thoughs for GPT-5 and AI agent market focused on security issues
-With the arrival of GPT-5, he believes the AI agent market will experience tremendous growth
There are two bottleneck in future AI technology:
1. computational technical issue (e.g., FHEML, ZKML, and more powerful LLMs)
2. Security issue
Security issue
-Security issue is concerning due to the potential risks with user authoritys (e.g., access to email, operating systems, etc.)
-Based on his experience, security issues are crucial for the advancement of AI
AI Agent
-With the arrival of GPT-5, he anticipates that Web3 AI agent projects will achieve significant success
-However, business models such as AI Auditing DAO, H2E (Hacking to Earn) could also emerge with potential

Bittensor
-Should check out s14 of bittensor They are doing prevention of attack for LLMs

📚 AI Project Research

Article on Heurist by Junho Kim, L Tony - Full article link: https://t.co/29RZzrvt5O

Short Summary
-We can classify Heurist into ‘Open source AI Model’, which can be expected as a web 3 AI model hub for dApps
-It will be difficult to compare the open source model with proprietary models that have Hundreds of M/B parameters
-Open-source models are expected to be in demand among developers who want to develop customizable models using few parameters
-If we can distribute and integrate Crypto-based models within Heurist and provide AI tools, it has potentials to become a hugging face in Web3
-Heurist is a L2 network developed with ZK stack for AI model hosting and inferences
-Consumers, Miners/Model Hosts, Model Creators, Application Integrators, and Validators are participants for the network
-Token scheduled to be released in the third quarter of 2024
-Competitive projects(Similar concept): Ritual, Neura, Nimble, Allora, Shinkai

Some more bullish points of Heurist
-One of a few Hyperstack(developed with ZK stack) project
-So zksync team could push it + Indirect benefits after zksync tge
-Clear tokenomics & token distribution → could show a good performances
-After watching the community for several months, can feel that the team is working hard
-Products are continuously coming out
-If the funding round is successfully completed, it will be quite good

[AI-meda Research @AIMedaResearch]

2024.04.16~2024.04.17 daily summary for public - 국문

🗣 Main Discussion

Heurist
-NFT 홀더들에게 1.75%의 토큰을 에어드랍하자는 프로포절이 나옴: https://t.co/d0Wj2nEd1c
-현재 nft fp는 2.5eth 정도
-초기 FDV는 지금 nft 가격으로만 본다면, 최소 200m을 예상

https://t.co/tZITgfR2nu - 토크노믹스
1. 고정된 Supply : $800m
-500 million $IO가 런칭하면 분배될 것
-300 million 물량은 Supplier와 Staker에게 시간마다 보상으로 지급하는데 사용
2. Suppliers, Stakers에 대한 Hourly rewards
-20년에 걸쳐서 분배됨
-첫 해에 8%가 분배되고, 매 달 분배량이 1.02% 감소함
3. 소각 메커니즘
-$IO는 프로그래밍된 코인 소각 시스템에 사용될 것
-IOG 네트워크에서 https://t.co/hVG9zVyDoa에 의해 생긴 수익은 $IO를 구매하고 소각하는데 사용될 것임
-$IO 소각 메커니즘은, $IO 가격에 기반하여, 소각될 물량을 결정

Quantum Cat
-모든게 CatVM을 위한 빌드업일 가능성이 있음 / reference: https://t.co/ZteCzGpqaq

[AI-meda Research @AIMedaResearch]

2024.04.16~2024.04.17 daily summary for public - English

🗣 Main Discussion

Heurist
-Proposal for 1.75% token airdrop to NFT holders: https://t.co/d0Wj2nEd1c
-Current nft floor: 2.5eth
-Initial FDV should be at least 200m maybe at this nft price

https://t.co/tZITgfR2nu - Tokenomics
1. Fixed supply : $800m
-500 million $IO coins will be distributed at launch
-300 million will be emitted and paid to Suppliers and their Stakers as Rewards each hour
2. Hourly rewards to Suppliers and their Stakers
-Released to Suppliers and their Stakers hourly over 20 years
-Starting at 8% in the first year and decreasing by 1.02% per month
3. Burning mechanism
-$IO uses a programmatic coin burn system
-Revenues generated by https://t.co/hVG9zVyDoa from the IOG Network are used to purchase and burn $IO
-The $IO burn mechanism adjusts the amount of $IO to be burned based on $IO’s price

Quantum Cat
-All of them could be a build up for CatVM / reference: https://t.co/ZteCzGpqaq