Lobster Daily 🦞

Agentic Marketplace from Grok 👨‍💻🤔

https://x.ai/bot/marketplace

upd: Open AI moves away from Cursor because “Elon Musk's companies are violating contracts”

https://openai.com/index/our-decision-on-cursor-following-its-acquisition-by-spacex/

llm watermarking is utilizing #synchID can be easily solved
just by tokens resampling using small optimized open weight models or cheap chineese cloud models

misAnthropic is basically incentivizing its own users to apply distillation for its own products rn

best technical explanation video:

https://youtu.be/FnAqruxx-QE?si

Lobster Daily 🦞 pinned «more attention to:»

Lobster Daily 🦞 pinned a photo

It’s been a while. LobsterDAO was, and still is, a valuable source of insights, experiences, and debates for the Web3 and AI developer communities. Although, the authors of this digest tend to be more interested in agentic engineering and its applications than in Web3 and DeFi.

There is a new digest announced by LobsterDAO:

https://t.me/lobsters_chat/583280

Please follow it for updates.

This daily digest will be refocused on agentic AI and its applications in coding, automation, agentic e-commerce, cybersecurity, and research.

There is a vote below on which topic you think we should focus more attention on:

LobsterDAO 🦞 — Daily Recap — 23 Jul

1. BitMEX shuts down — discussion of Arthur Hayes era, exchange history, and predictions ex-BitMEX team will build new perp DEX
Link to discussion

2. AI models for onchain data analysis: Claude vs GPT vs Gemini vs Fable, agent swarms, and subscription cost economics
Link to discussion

3. Uniswap V4 hooks fail to gain adoption due to complexity and poor static analysis; users also frustrated by missing zap functionality for LP deposits
Link to discussion

4. Synchronous composability between rollups: Norswap questions technical feasibility, Gnosis EEZ team shares workshop resources and real-time proving approach
Link to discussion

5. Rabby wallet removes custom RPC button and enables data analysis by default — users frustrated, migrating to Ambire; Ambire team actively gathering feedback
Link to discussion

6. Robinhood CEO Vlad Tenev's Twitter hacked — hackers launch coin generating $10M volume; discussion of phishing techniques, SIM swapping costs, and why people buy hacker coins
Link to discussion

7. New member asks about building a memecoin launchpad — met with skepticism and comparisons to pump.fun draining industry liquidity
Link to discussion

DeFi peoples, brainraping time.

When you consider the risk of a protocol or an asset to deploy into, it comes through different angles:

1. code review: codebase itself (architecture, its simplicity, audits, bug bounties, test coverage, formal verification, etc)
2. liquidity of an asset: how liquid is a stablecoin, for example, when can you withdraw, how long to wait until maturity, etc. this isn't a risk in a hack sense but a risk nonetheless
3. leverage ratio: if is a lending protocol or something that underwrites loans/mints - the question is also how much is it looped. the more looped it is, the riskier sometimes an asset/protocol is (not always I guess). this can be similar to point (2) though
4. dependency on third-party things: oracles, keepers, curators, or any other jobs (onchain or offchain) that can influce the behavior of your position. influence, not "own" per se
5. admin access controls: admin permissions in contracts, who can mint or freeze, who can block assets or add new collaterals, etc. defiscan does that, for example
6. legal aspects: this is not per se point 5, but is an extension of it. what is the credibot-debtor relationship and how could that mess up your position even if no hack was done to your assets directly (like kelp L2 situation with bridge dependency, or how it was with Resolv to an extent where it wasn't a crystal-clear choice)

Questions:

a) What are the 6/7/8 angles you can add to the list?
b) What are the services today that review these things? DeFiscan for (5), DeFi Daver / Chaos Labs / a few other for (2 or 3)... what else? What about the other pointers?

"Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans"

https://www.anthropic.com/glasswing

Agentic Love:

https://t.me/lobsters_chat/567728

Adding some perspective - L2s account for ±2/3 of all chain-level fee generation within the ETH ecosystem today.

But fee generation as a whole, has been on a decline within the ETH ecosystem. Tx fee as a model itself may come to question, unless ETH decides to rent-seek on liquidity moat (which is a core pitch made in the doc above). FWIW, that is a justifiable outcome given L1 security and uptime performance so far

Edit: what i'm trying to say is - both L1 and l2 may undergo a period of business model innovation- easier at the l2 level