# DeFi — X 热门讨论 (2026-09-26 16:48 UTC)
## @XiaoZhi_BTC (小智🎒) · 09-26 13:55 · ♥81 ↻1 💬96 一般提到「股票能当抵押」，脑子里先蹦出来的画面是：券商给你开个融资口子，拿持仓去借点钱。
但 @Backpack 这回做的，不太是同一个东西。
它把「真实美股持仓」和「链上交易账户」塞进了同一个统一保证金池。股票不用卖，就能在同一个钱包里借 USD、跑永续、做现货保证金——而且这些仓位共用一份抵押，不像别家非得把现货、期货、保证金几个账户拆开摆。
这点区别看着小，其实挺关键。
底层是因为它手里的股票是「真股权」：每一股对应美国托管里的实股，按 UCC 第八条算证券权益，走 ACATS/DTCC 那套清算托管。不是拿价格做个合成敞口就算了。
更妙的是，这些股票还能在 Solana 上代币化——自己托管、7×24 交易、能直接接 DeFi。也就是说，押在保证金系统里的那批持仓，本质是个链上原语，能伸到更宽的 Solana 生态里去，而不只是券商账本上的一行数字。
这层传统券商复制不了。
能当抵押的标的，已经从最早的 MU、SNDK 扩到 17 个：TSLA、NVDA、AAPL、GOOGL、META、AMZN、AMD 都在名单里。量也起来了——Backpack 一家就喂出了 Solana 代币化股票 30 天 33 亿美元成交量里的 11 亿，占三分之一；$GRND 上线两小时干了 1410 万，比它在纽交所当天的量还高出三倍。
所以这事真正该问的，不是「我能不能拿特斯拉股票借点钱」。
而是：一个把持牌经纪商和链上保证金引擎揉进同一个钱包的账户，会不会就是下一代金融账户该有的样子。
@Backpack_CN @BackpackOnchain > 引用 @Backpack: Do more with your stock portfolio on Backpack.
▸ Borrow USD against your holdings ▸ Trade perpetuals and spot on margin using stocks as collateral ▸ Hedge positions without selling your shares
17 stocks and ETFs supported as collateral today. More coming. https://t.co/EhWF8T4e5E https://x.com/XiaoZhi_BTC/status/2103845797251199227
## @Utpal733128 (Utpal) · 09-26 14:40 · ♥73 ↻17 💬65 One of the biggest problems with Web3 identity is that your reputation is often trapped inside the platform where you earned it.
You might have years of activity on an exchange. A strong credit history somewhere else. Verified credentials from another service. Or meaningful contributions across different platforms. But when you move to a new application, you often start from zero.
That is the part of @primus_labs I find interesting.
What if your existing online activity could become verifiable without forcing you to expose the underlying data?
Instead of saying: Trust me, I have the history.
You could generate a proof that says: This specific requirement is true.
The application can verify the claim without needing access to your entire account.
That creates a different model for digital identity.
Your data can remain where it already exists. Your proofs can move between applications.
And your reputation doesn't have to be rebuilt every time you enter a new ecosystem.
This becomes especially interesting as DeFi, AI agents, social platforms and real world services start requiring more reliable user information.
The future of identity may not be about putting everything onchain. It could be about making the right facts portable while keeping the underlying data private.
Prove the signal. Keep the source protected.
That’s the direction I’m watching with Primus. > 引用 @Utpal733128: Robotics scaling is starting to look very different.
@axisrobotics is building toward a world where robot training is not just about collecting more raw trajectories, but about turning experience into reusable skills.
Three ideas stand out:
• Embodied RSI: deployed policies generate real world data that can train stronger successors, with success improving from 22% to 52%.
• Smarter data selection: a lightweight proxy model evaluates trajectories by utility and coverage showing that a carefully filtered dataset can outperform the full pool.
• Composable expertise: expert models can reach near 100% task success for roughly $5–$10 in compute and then be chained together for longer horizon behaviors.
The bigger picture is interesting.
Instead of scaling only the amount of data, Axis is scaling a growing library of robotic capabilities that can be produced, reused, composed, and distilled into more general purpose models.
That could make every successful robot interaction another building block for the next generation of physical AI. https://x.com/Utpal733128/status/2103857151978717545
## @davidgokhshtein (David Gokhshtein) · 09-26 15:48 · ♥89 ↻20 💬28 When $NIGHT runs like $ZEC, nobody should act surprised.
I’ve been saying privacy was going to matter again and $ZEC is showing you that right now.
What I like about Midnight is privacy isn’t just the asset. It’s built into what people can actually build on top of it. Privacy in smart contracts, selective disclosure and applications that can actually be used outside of crypto.
Then you have the Cardano side of it and everything happening around Bitcoin DeFi and bringing more BTC liquidity into that ecosystem.
I’ve spent a lot of time looking into Midnight and talking directly with the people building it.
There’s a reason I keep talking about $NIGHT.
https://t.co/dLTHRKjTmh https://x.com/davidgokhshtein/status/2103874243662434361
## @Pluto_Crrypt1 (Plchoo | 火币余币宝13%) · 09-26 13:05 · ♥86 ↻1 💬49 同样拿着 fxUSD，不同的人想要的东西其实差很多。
有人只想有一笔能随时使用的稳定币；有人愿意把它放进 fxSAVE 获取收益；还有人更在意未来某段时间的收益怎么定价。@protocol_fx 相关资产进入 Pendle 后，这些需求开始有了不同的表达方式。以 fxSAVE 对应的市场为例，收益权和本金相关头寸可以分别交易，用户不必都挤在同一种持有方式里。
我觉得这比单纯宣布“又多一个 DeFi 集成”有意思。稳定币能走多远，靠的往往不是让所有用户做同一件事，而是有人拿它储蓄，有人拿它交易收益，有人把相关资产用于抵押。每多一种真实用法，fxUSD 就多一个留在链上流转的理由。
当然，Pendle 的隐含收益率会变化，PT、YT 的价格也不是只看一个 APR 就能判断。到期时间、市场深度、提前退出的价格，都可能影响最后的结果。没弄懂这些规则之前，把它简单理解成“固定高收益”，很容易算错账。
所以 fxUSD 到 $85M 后，我想观察的是需求有没有开始分层。铸造量增加是供给侧的故事；不同用户愿意围绕它做不同安排，才说明生态真的在长。
数字已经有了，接下来就看这些用途能承接多少真实资金。 > 引用 @protocol_fx: $85M fxUSD in circulation.
$80M felt like a milestone. $85M feels like a trend.
fxUSD supply is still climbing, and nothing is being paid to make it happen. People want a dollar that stays a dollar, and they're minting it.
Hold it. Earn on it in fxSAVE. Borrow it against ETH or BTC in fxMINT. https://x.com/Pluto_Crrypt1/status/2103833432468156470
## @bi_9527zx (DeFi狙击手 | Ai🕊️) · 09-26 14:46 · ♥76 ↻1 💬23 我艹睡不着了睡不着了，
100倍，100倍，哈哈哈！ https://t.co/QJHi98c5Fv https://x.com/bi_9527zx/status/2103858698670203055
## @DamiDefi (Dami-Defi) · 09-26 15:34 · ♥75 ↻7 💬8 This is insane!
Grok Bot took 7m 40s. This AI assistant did it in 22 seconds.
Cerebras built a faster Grok Bot-style assistant using Qwen 3.8 27B + Pi, cutting the recorded dinner-reservation task by roughly 19×.
The biggest gains came from:
→ Running independent checks in parallel → Saving website procedures as reusable skills → Reducing tool calls by 80%+ → Using much faster model inference
The real lesson:
Agent speed isn't just about the model. It's model + harness + tools + memory + workflow design.
Faster agents could make autonomous AI feel less like waiting for software and more like delegating to it. > 引用 @cerebras: We built a 20x faster Grok bot https://x.com/DamiDefi/status/2103870712708788238
## @0xBispo (Defi Priest) · 09-26 14:18 · ♥79 ↻7 💬1 🚨 IT’S LIVE 🚨
Ready to turn 45 mins of playing into a shot at 200K $PLONK? 👀
The #PLONK Bounty Quest is LIVE NOW.
🎮 Tutorial → gather → craft → chop → mine → fish → fight 🏆 25 winners 💰 200K $PLONK each 🔓 200K unlocks Tier 2 Market access
And yep… you actually have to PLAY. 😂
⚠️ NEW PLAYERS ONLY You MUST use my referral before creating your first character:
👉 https://t.co/A2MAqVjL6m
Best #P2E in Solana rn https://x.com/0xBispo/status/2103851753057751474
## @Lowdef1 (Low DeFi) · 09-26 12:56 · ♥73 ↻1 💬0 The 10Y Treasury yield is becoming one of the first charts I check now. 👀
With yields around 5%, I’m finding it harder to look at BTC, stocks or gold in isolation.
Higher yields can mean: 📈 More pressure on stock valuations ₿ A tougher liquidity backdrop for crypto 🥇 Higher opportunity cost for gold 💵 More support for the dollar
But none of these are automatic trades. Inflation, growth, positioning and geopolitics can change the reaction quickly.
So my workflow has changed.
10Y first → then BTC, stocks, gold and USD.
I keep those markets open on BingX and adjust my strategy as the macro picture develops.
Which market do you think reacts most noticeably when yields move? https://x.com/Lowdef1/status/2103830948592697499