Mohamed Fouda (@fouda) • Hey
Contributor @alliancedao, Venture Partner @VoltCapital, Engineer and Researcher
Publications
- Today, we launched Alliance Research Corner, a series of focused technical discussions on crypto technical topics.
Our goal is simple. We want to help the best founders find their next startup idea.
The videos are short and focuses on one topic and the related startup ideas.
Give it a listen and let us know what we should discuss next
https://www.youtube.com/watch?v=tTUlyxjkyNA
- My new article discusses the different approaches to scaling app Rollups including
1. Horizontal scaling, i.e., launching multiple Rollups serving the same app
2. ZK state channels: launching a dedicated state channel for each game on the roll-up that settles back to the L2S
3. Changing the Rollup execution: builders are using WASM, SVM, Move, and even Linux as an execution env for their app Rollups
I would love to speak with developers building these applications
https://medium.com/alliancedao/how-to-scale-app-rollups-ad2313576efa
- We just wrapped a 🔥 Alliance demo day
Eight companies have graduated today from the Alliance ALL11 cohort.
These founders are part of Alliance Dao and part of a unique community of top founders in the space.
All future builders can become part of this community. ALL12 applications are open until November 29. Apply now at Alliance.xyz
https://twitter.com/jacqmelinek/status/1714670721467654255
- The new Good Game podcast just dropped with the one and only @stani.lens
Imran and Qiao discuss with Stani the Lens V2 vision and how decentralized social will work in the future
YouTube: youtu.be/xorv83lN8K8
Spotify: spoti.fi/3t1A0vt
Apple: apple.co/48ix0Lk
https://www.youtube.com/watch?v=xorv83lN8K8
- A year after the Merge, the amount of staked ETH has doubled, and the supply of liquid staked tokens (LST) has more than doubled. The effects of both trends on the Ethereum network will be a major topic of discussion in the next few months.
On one hand, the trust in Ethereum’s security is high and the staked ETH amount is constantly on the rise. One the other hand, higher amount of staked ETH means higher # of validator nodes straining the network and can affect its reliability!
Further, LSTs offer staking rewards without the burden of locking liquidity. Hence, they reduce the network’s security budget. Other PoS protocols, e.g., Cosmos, began to limit LST supply to a reasonable percentage of the staked tokens. Ethereum can also adopt similar measures. However, with the current pace of staking, the Ethereum community can run into a crisis before such a solution is implemented.
For this reason, Ethereum researchers are proposing reducing the are new stakes can join he network, e.g., validator churn rate. This can be an effective solution to solve the issue of the # of validator nodes. The LST cap problem will be harder to fix and will require a larger social consensus.
- secrets (revealed)
- How do Lens folks feel about Friend.tech it seems as an experiment of hyper financialization of decentralized social. Any thoughts?
- Switch between networks automatically
Browse the web3 with ease when switching to multiple chains with Zerion Extension.
zerion.io/extension
- On-chain privacy will become a central theme in the applications/products that initiate the next bull run. The core of this theme is enabling multi-asset privacy. Zcash and Monero failed because only a single asset is private. Users want stablecoins and blue-chip assets and privacy should target this needs.
Manta and Namada are two projects that aim to bring privacy to existing chains by creating chains with specialized zk circuits that can accept any asset (fungible and non fungible) and create shielded assets that can be transacted privately. There are differences in the implementation between both but the architecture is similar.
Manta launched a Polkadot parachain (Manta Atlantic) and is launching an Ethereum OP rollup (Manta pacific) to allow devs to use essential zk primitives (called Universal Circuits) to develop privacy-centered apps and reduce their time to market.
Similarly, Namada is launching a Cosmos comaptible chain with a single multi-as @ set shielded pool to provide privacy to both Cosmos and Ethereum assets. There are multiple issues to consider though such as bridging risk and the privacy of posting data to an external DA such as Celestia.
Other than ZK tech, fully homomorphic encryption (FHE) seems to be approaching. The Zama.ai team is developing an FHE-compatible EVM (feEVM) that can enable computing over encrypted data. The technology is still early with limited functions and limited variable precision (INT32) but can enable unprecdended use cases.
- My friend Will Robinson (our gaming expert in Alliance) has just released an article detailing a technical architecture that can bring games such as World of Warcraft on-chain!
The secret is to intelligently use ZKPs to take parts of the game play offchain. When a player or a group of players enter a dungeon in WoW, they can so the game play offchain (in a provable state channel) and then prove their honest gameplay on the shared game state. Many players can do the same concurrently bringing 1000x scalability improvement to on-chain games.
Also great probs to the teams building this tech including the Dojo team!
[Unblocking On-Chain Games Pt. 5: Bat-Channels](https://medium.com/alliancedao/unblocking-on-chain-games-pt-5-bat-channels-f24a3845e465)
This is Part Five of a series exploring constraints at the intersection of blockchains and games. The objects of inquiry are “autonomous…
- The new Bookmark feature is live on @lenster.lens, allowing users to add Bookmarks to their favorite posts. There's also a new 'Not Interested' option for content that doesn't match your taste.
Dive into the dev docs:
https://docs.lens.xyz/docs/bookmarks
https://docs.lens.xyz/docs/not-interested
- We have been planning this episode for a while. ZK is one of my favorite topics and in this episode we went really deep to discuss
- the fast progress of ZK tech (a hint: more than 30x annually)
- how cryptocurrencies were a major factor in developing the ZK tech that will affect many use cases in Web 2 and Web3
- How ZK accelerators can give you up to 100x boost and why this may still not be enough
What the episode here
Youtube: youtu.be/-vI-bSR0kY8
Spotify: spoti.fi/3NYq5yz
[What are Zero-Knowledge Proofs and Why It’s Important?](https://www.youtube.com/watch?v=-vI-bSR0kY8)
Imran, Qiao, and Mohammed invited David again to delve into the world of Zero-knowledge proofs. Join us as we demystify and uncover why they have become incr...
- The new Good Game episode just dropped. I had a 🔥 discussion with Qiao, Imran and David on intents, what they are and the hype around them.
We also discussed the details of Anoma and whether it can work with Ethereum or not
[Is Intent-Based Architecture a Major Breakthrough Like Bitcoin & Ethereum?](https://www.youtube.com/watch?v=7g5GNq_jXR8)
What are Intents? Is it going to be a major breakthrough or it's just another buzzword? Imran, Qiao, and Mohamed invite David Ma to answer these questions an...
- The deadline for Alliance ALL11 cohort applications is tomorrow!
Although builders can learn a lot about Alliance by watching the accelerator’s graduated companies and how well they execute and build great products. It’s still helpful to explain how Alliance supports these companies
1. Our priority is to be engaged and available to help. Our calendars are open for Alliance founders to book mentorship time
2. The accelerator core team members are experts in their respective fields and each has years of experience in Web 3 space.
3. You not only get help from the Alliance team but from successful founder who graduated the program and leading successful companies in the space.
This is just a small subset of the benefits of joining the Alliance community.
Personally, I look forward to working with builders is the ZK space, AI × Crypto, appchains and the intersection of Intents, AA and MEV.
Don't miss the window and apply today!
[Accelerator Application | Alliance](https://alliance.xyz/apply)
Apply to our accelerator program and join the leading network for Web3 startup founders.
- Some brilliant stone stacking skills
- Recorded a new Good Game podcast episode today on the new fad in crypto land "Intents" 😂
Anoma announced $25M raise last week so it was appropriate to dissect what are intents, what are intents solvers how Anoma and Suave work. It was a great discussion.
What do you think of intents?
- At Alliance we work with founders to build world-changing companies of the future.
I am personally stoked to work with ALL11 founders. You still have a chance to join an awesome group of founders.
The deadline to apply for ALL11 is June 14th.
Submit your application today
[Accelerator Application | Alliance](http://alliance.xyz/apply)
Apply to our accelerator program and join the leading network for Web3 startup founders.
- A new 🔥 episode on the Good Game podcast. I joined Qiao, Imran and Will to discuss what is happening in Solana land.
We discuss compressed NFTs, Solana Phone and other catalysts that make us believe that Solana is taking off!
🔊: bit.ly/3IBSzLt
📺: bit.ly/3IuZMx6
[Is Solana on track to follow Ethereum's Success Story? | GG Research](https://www.youtube.com/watch?v=Es863jKCPRU)
Imran and Qiao enlist the expertise of Mohammed Fouda and Will Robinson for an in-depth exploration of Solana.No BS crypto insight for founders.Timestamps00:...
- Last week, we shared how @lensprotocol built Momoka, its innovative scaling solution, using Bundlr
Let's dive into what that means 👇🏾
Momoka uses Bundlr for scalable storage and ordering, and Polygon for smart contract interactions
One of the most significant challenges of this hybrid approach was maintaining transaction order across the two different systems
Lens solved this problem with a simulation solution that guarantees the real-time chronological sequence of user transactions, regardless of whether they are posted to Polygon or Bundlr 🚀
Bundlr as a storage and ordering layer is 1000x more cost-effective with 700x lower latency. Since launching last week the average Momoka transaction cost is only $0.0005 🤯
Bundlr's ordering service is fully permissionless, and is detailed in our docs 🤝
https://docs.bundlr.network/overview/services/ordering
Stay tuned for one more exciting announcement from Bundlr and Lens. In the meantime, dive into our blog post on how Momoka was built using Bundlr 🌟
https://bundlr.mirror.xyz/oDs2P54Esh9Tmi8OIL7KGVVXH_hZAwkW8JUC4cVrCL4
- We had a very interesting discussion on how AI and Crypto intersect on @Bankless.
We have discussed how AI agents can become autonomous and participate in on-chain games and decentralized social media. Crypto actually provides the perfect payment and execution rails for autonomous AI agents.
We also discussed how Crypto can solve problems like deep fakes and privacy of data in ML models.
Give it a listen and let me know your thoughts!
[AI and Web3 | Mohamed Fouda & Qiao Wang of Alliance](https://www.youtube.com/watch?v=FyEcfgbCrDw)
AI is exploding into every facet of the internet. The convergence of Crypto and AI is inevitable, and we bring on Qiao Wang and Mohamed Fouda to discuss the ...
- A really fun podcast just I recorded with Qiao Wang and Imran Khan was just released.
Crypto + AI: Uncovering the Next Unicorn
In this episode we discuss:
- A vision for decentralized AI
- How crypto can solve the drawbacks of AI
- Why AI Agents can be the future of crypto
Let me know what other use cases you think will emerge by merging forces between AI and Crypto!
🔊: bit.ly/3p03gRg
📺: bit.ly/3VmkELU
[Crypto + AI: Uncovering the Next Unicorn | E16](https://open.spotify.com/episode/7lyXRWutr03BLPRZkW0JOv)
Listen to this episode from Good Game on Spotify. Imran, Qiao, and Mohamed Fouda deep dive into the convergence of Web3 and AI.No BS crypto insight for founders.Timestamps (00:00) Intro (00:42) Welcome to Good Game! (01:34) Fouda and Qiao’s backgrounds in AI (06:13) ChatGPT is creating 10x developers (09:42) A vision for decentralized AI (24:57) Drake AI music controversy (34:15) Sam Altman's Worldcoin proving human identity (41:14) How drawbacks of AI can be solved by crypto (48:38) AI Agents can be the future of crypto (1:02:56) Identity war is coming in a post AI worldThe Convergence of AI and Web 3 Startup Ideas: https://alliance.xyz/ideasMohamed Twitter: @mohamedffoudaGood Game Podcast: A podcast for crypto natives. Find us at goodgamepod.xyz and follow us on Twitter @goodgamepodxy
- In my new article with Qiao Wang we discuss the synergies between Web 3 and AI and how both technologies can converge to enable new use cases and address some of the existing challenges for AI systems.
Our thesis is that Web3 and AI are culturally and technologically compatible with each other.
Web 3 allows AI to flourish thanks to its permissionless nature. Specifically, if we view blockchain as a network, then we envision AI to dominate the edges of the network. This applies to all sorts of consumer applications from social media to gaming.
Building on this mental model we discuss several use cases that AI can enable in Web 3 including how AI agents can revolutionize on-chain gaming, decentralized social media, DeFi and even infrastructure.
On the other hand, Web 3 can help AI address some of the challenges around performing inference on private data, building of proprietary datasets, and centralization.
Zero-knowledge proofs, PoPW networks, and decentralized collaboration are all Web 3 ideas that will play a significant role in the future of AI.
Web 3 can also the misinformation problem resulting from AI-generated deepfakes. Finally, Web 3 networks are idea environments of AI agents such as AutoGPT. These AI agents will have their own wallets to receive payments for the work they do and even hire and pay other AI agents. They can also commission resources they need to execute certain tasks whether these resources are decentralized storage or decentralized computing. They can even commission these resources to scale their own performance!
[The Convergence of AI and Web 3: Opportunities and Challenges](https://medium.com/alliancedao/the-convergence-of-ai-and-web-3-opportunities-and-challenges-45c28105e513)
By Mohamed Fouda and Qiao Wang
- I joined Qiao Wang and Imran Khan on the Good Game podcast to discuss the current roll-up space and how it can evolve.
We discussed
- Optimistic rollups vs ZK rollups
- Current state of rollups and tradeoffs
- What we're seeing in dev activity
Spotify 🔊: bit.ly/3zSZ47W
YouTube 📺: bit.ly/40aMNXq
[The L2 Wars: Which Optimistic or ZK rollup should Web3 Founders choose? | E15](https://www.youtube.com/watch?v=KXCVdOT594Q)
Imran and Qiao bring on Mohamed Fouda to talk L2s.No BS crypto insight for founders.Timestamps00:00 Intro00:34 Welcome to Good Game!00:48 Welcome Mohamed Fou...
- The rise of chatGPT and LLMs has revived the narrative of decentralized computing but with a focus on AI.
Here, I am presenting an antithesis on the viability of decentralized computing for AI computing, especially for training!
Despite being an advocate of decentralization, I believe it needs to serve a purpose that cannot be achieved otherwise.
I am not convinced that decentralized computing offers a 10x better compromise than centralized solutions
Feel free to correct me if I am wrong!
Let's start with the basics:
LLMs are massive models with billions of parameters to train. GPT-3 has 175 billion params. GPT-4 is speculated to have 1 trillion params.
The no of operations (FLOPS) to train these models is 10^12 x the number of model parameters. GPT-3 requires 3.14e23 FLOPs for training.
Specialized hardware, i.e., GPUs, is needed to perform these computations.
If you think you've these GPUs in your gaming or mining rig! The answer is No!
AI GPUs are special GPUs that are optimized for AI and have special components, e.g., Tensor Processing Units (TPUs).
An example is Nvidia's A100. It has a beast chip with 54 billion transistors. It can perform 300+ TFLOPs.
Even with such impressive performance, training GPT-3 with a single A100 takes ~ 33 years!
Instead, you can have 1000s of GPUs working together to perform the training.
Nvidia is also developing the H100 AI GPUs which have a unique computing element called Transformer Engine that target the specific AI architecture that LLMs use. The expected performance boost is ~ 9x
Decentralized AI computing proponents say that a decentralized network of GPUs can do this task even faster!
That is wrong!
The details of this how AI computing works matter a lot in this discussion.
First, we established that AI GPUs are different from mainstream gaming GPUs. AI GPUs use state-of-the-art semiconductor fabs (7 nm and 5 nm) which is in high demand creating a semiconductor shortage. Unless you are a powerful entity you cannot secure these GPUs.
Second, for optimal performance, all the GPUs need to be identical, i.e., homogeneous computing, otherwise, stronger GPUs have to wait for slower GPUs before they can exchange data and continue computing meaning that the system will work at the speed of the slowest GPU. This means that for a decentralized computing network to be efficient, all participants should have the same or at least comparable GPUs.
Finally, AI GPUs have to exchange data all the time. The latency of data exchange is critical. That is why Nvidia builds specialized Pods with multiple GPUs connected by fast interconnect (NV-Link) to reduce this latency. If you move these GPUs apart, like in a decentralized network, the latency will go through the roof killing the overall system performance.
For all these reasons, decentralized computing is not really a fit for AI training.
The performance hit is unlikely to justify any reduction in cost you can get via decentralization. Instead, our efforts should be directed into more impactful areas of AI such as
- collaborating to build new, better, and open models
- collaborating to create better data sets for the open-source models
- Attention Web3 founders! It's that time of the year again.
Early admissions are open for the next Alliance Accelerator Cohort (ALL11). We're seeking diverse and innovative minds to join the community. Open to all early-stage Web3 startups.
Apply by April 26, 2023 https://alliance.xyz/apply
- 📢Attention Web3 founders! It's that time of the year again.
Early admissions are open for the next Alliance Accelerator Cohort (ALL11).
We're seeking diverse and innovative minds to join the community.
Open to all early stage Web3 startups.
Apply by April 26, 2023 👉
- I am a big believer in the potential of ZKPs to transform our world and enable a new trust paradigm between companies and their users.
In my new article with Qiao Wang, we discuss how Web 3 builders can leverage ZKPs to improve their products and achieve a strong PMF.
We start by discussing how ZKPs work and how the separation of the proving and verification parts enable new dynamics that can expand the trust-minimization ethos from Web 3 to Web 2.
We then discuss the parameters of ZKP systems and how they affect the product performance. We use zk rollups as an example to show how the ZKP proving time determines the rollup withdrawal delay and how proof size and verification cost affect the rollup tx fees.
We then introduce a framework to help developers asses how ZKPs can benefit them the most. We discuss the available tools to help devs efficiently integrate ZKPs in their products and the effort needed for each use case.
[How to Leverage ZKPs as a Web3 Builder](https://medium.com/alliancedao/how-to-leverage-zkps-as-a-web3-builder-ae504783973d)
By Mohamed Fouda and Qiao Wang
- GitHub active project
- The launch of Arbitrum's Orbit and Optimism's OPstack before that along with the proliferation of L2-as-a-service providers represent a major competition to Cosmos
What works in Cosmos's favor is
- Simplicity and modularity of the Cosmos_SDK
- IBC bridging is seamless and has top-tier security guarantees
- Easily customizable appchains
What works in Ethereum's L2s' favor
- Shared security with Ethereum (no need to recruit validators)
- Programmable out of the box (EVM/Solidity)
- Larger dev ecosystem
- Abundant service providers for the EVM
Cosmos also benefits from Sovereignty by internalizing gas fees and MEV activity, e.g., Osmosis, but some L2s will find ways to achieve similar benefits
- 📢 Last call for applications!
Thanks to all the incredible startups that applied
We’ve already received a record-breaking number of submissions
We can’t wait to welcome the next batch of Web3 founders 🤝
Apply now: alliance.xyz/apply
- 📢 Applications to Alliance’s upcoming accelerator cohort close on Jan 16.
Open to all early stage Web3 startups.
Apply Now!
- Planning a trip to Casablanca in the next few weeks. Any recommendations for what to do, places to visit. Also open to visit other cities in Morocco 🇲🇦
- My most read article of 2022 was "The case for parallel processing chains" where I discussed why parallel processing chains are important, the different kinds of implementation, and how these implementation compare.
I am still bullish on how these chains can bring Web 2 performance to Web 3. Which of these chains are you the most excited about?
[The Case for Parallel Processing Chains](https://medium.com/alliancedao/the-case-for-parallel-processing-chains-90bac38a6ba4)
When looking at the evolution of blockchain technology, we can recognize a strong trend of new L1s that focus on parallel execution. The…
- 📣 Attention Web3 builders!
Regular Admissions are now open for Alliance Accelerator Cohort 10 (ALL10).
Open to all early stage Web3 startups.
Apply now: alliance.xyz/apply
- Woohoo!!
Just claimed fouda.lens 🎉🎉🎉