@Smore ★★★★☆

I've been doing some DD into $SODI on Robinhood Chain. 0xEc383A25b6434216eaA695c11c26B765BA061521 The idea is to borrow against tokenized stocks, then spend the borrowed USDG without leaving an obvious trail from your loan to everything you do with the money. In the first version, the position itself stays public. Hiding the position is a later stage. I can see the appeal. Onchain finance makes your collateral, debt and trading activity visible to anyone willing to look. Borrowing against something you want to keep is already a familiar idea. Giving people more privacy around what they do with the proceeds could make that useful to a particular group of traders. I’ll cut to the chase, what got my attention is that there's actual testnet code here to test, and this was very reminiscent of when I was DD’ing robinhood:0xdcd885f5c60b3ad49f6d58718696417314e8724d. There’s three types of code here. Mainnet, testnet, and code I had to fork. The borrowing side uses Morpho, which isn’t available on Robinhood’s public testnet, and thus I used a local mainnet fork to test what required Morpho. Mainnet has their swaps which is cool and all, but isn’t the main selling point of $SODI. So on testnet, the deposit-approval service, withdrawal relayer and AI gateway were all responding and operational. I traced existing transactions showing testnet USDG moving out of the privacy pool into a session wallet, followed by a successful stealth payment to a merchant address. Separately, I sent a fresh request to the free AI gateway. It correctly answered my question, showing there’s real inference layer attached to the project. The agent side gives the same payment infrastructure another potential use, although the full agent interface isn't live. There are functioning payment components. For the blockchain flow, I verified existing transaction receipts, but didn’t complete a new paid session or establish that the transactions couldn’t be linked through timing or other clues. The stock-backed borrowing product wasn’t deployed on public testnet yet, so I downloaded the code and tested it against a local copy of mainnet. I pulled the original repository onto my Windows server and ran 66 selected checks, out of which 65 passed. That included 40 tests on borrowing, yield and deployment checks, eight solvency-proof verification checks, and 11 burner checks. The test that failed was inconsequential in the grand scheme of things and I’ll forward it to the dev for him to review. Granted I’m not a professional auditor, but these tests show there is concrete work behind the scenes. Usually the “coming soon, waiting for audit” that is on the website would give me pause, but given that the actual code is written and works, you know the actual hard part, I believe that they’re waiting on an audit to be finished. The borrowing code could borrow, repay, enforce limits and handle liquidation in a local copy of mainnet. The verifier accepted valid proof fixtures and rejected altered proofs or inputs. Some privacy-exit paths use mocks, so this wasn't a complete end-to-end privacy test, but I’m happy with what I found. There are working testnet payment components too. I verified fresh session addresses, stealth merchant payments and recycling leftover funds. I verified existing transactions and got a correct answer from the free AI gateway. Then there's the token. The fees are 2.8%. The stated allocation is 1.5% toward buying and burning SODI, 0.8% toward the borrower backstop, 0.2% for development and 0.3% for Pons. While team is handling its allocation manually during launch, automatic contracts are written and have working fork tests, but aren’t deployed on public testnet or mainnet yet. I also traced and confirmed that the burns are real. As of writing this, 4.65% of $SODI has already been burnt. The longer-term thesis is that useful private payments and borrowing bring customers, and customer activity eventually funds token purchases. Operators would also need to bond SODI under the proposed design. Those links still need to ship. Right now, the demonstrated burn activity is connected to token trading, not established production lending revenue. The devs GitHub is also 8 years old with a very very busy history of working on private projects. Like every month for the past 8 years there’s a ton of activity. Github also says that he previously worked at FlashBots and Aztec (Privacy based ETH L2). My reservations are straightforward. The testnet needs to ship to mainnet. Stock-backed borrowing isn't publicly deployed, and neither are operator bonding and slashing. There's no completed independent audit, adoption is unproven, and privacy can still leak through timing and amounts. But the fact that the testnet code is there, AND works, has my balls tingling. Feels like the robinhood:0xdcd885f5c60b3ad49f6d58718696417314e8724d trade all over again. Higher it goes! PS: While I was writing this they’ve launched the inference gateway on mainnet, followed by a console, chat interface and hosted MCP server. I checked that the new interfaces respond and payment quotes now request real mainnet USDG. That gives the AI side a route to paying customers, although I haven’t completed a paid request myself. The privacy pool and stock-backed borrowing are still separate milestones, but the fact that they're shipping makes me bullish.