psykeeper 𐁉

Founder @Saffron 𐁉 | Building DeFi fixed yield | Research focused | Ex pro-gamer

[psykeeper 𐁉 @psykeeper]

Let's talk about the high-level mechanics of Saffron Uniswap Vaults (release date TBD).

♨️Vaults transform Uniswap yield into fixed income
♨️LPs on the fixed side are paid up-front
♨️Variable side depositors earn multiplied yield on single-asset deposits

Read more below:

[psykeeper 𐁉 @psykeeper]

@srinigoes Check out fixed/variable yield for LPs on @Saffron (private beta soon)

Below is a screenshot of the $LIT pool we're testing

[psykeeper 𐁉 @psykeeper]

Saturday morning @Saffron post

[Dalton (Analyze & Optimize) @Outdoctrination]

Saffron restores motivation + spark for life in clinical trial.

Anhedonia is that feeling of meh. I don't care. Whatever.

Yet adding on 30 mg of saffron reduced anhedonia substantially over antidepressants alone.

This coincides with increases in dopamine neurotransmission and BDNF signaling.

[GEE-yohm LAMB-bear @guil_lambert]

Can't wait to get rekt on impermanent loss on my SPCXx-SPYx LP position when this launches.

[psykeeper 𐁉 @psykeeper]

Coinbase System Update in two weeks and my bet is: either AI or RWAs are the headline, or both

[Brian Armstrong @brian_armstrong]

It’s already been 6 months since our first System Update, so it’s time for another round of product announcements.

Our team has been heads down building - we’ve got lots to show you.

See you in two weeks! #SystemUpdate

[psykeeper 𐁉 @psykeeper]

@Brickken @Finextra @Ludovico__Rossi 🫳

[psykeeper 𐁉 @psykeeper]

@golieth_io *citation needed

[psykeeper 𐁉 @psykeeper]

@sir4K_zen I haven't tried it myself, but your comment inspired me to look into this, and I found that @stripe is probably best at this with their Stripe Treasury cards feature. Here is a mockup of the setup architecture:

[psykeeper 𐁉 @psykeeper]

Despite being a popular buzzword on Twitter and The Most Hyped Thing Lately, AI agents seem to be suffering some declining statistics and *gasp* charts going down and to the right. What explains this glaring discrepancy?

The most valuable thing an agent can do is anything you could do yourself. In that case, why haven't we seen payments taking off in the agent space?

Agents are still too hard to use securely.

To make an agent secure enough to handle payments, a lot of upfront logistics and programming work is required. This requires both spare time and specialized expertise.

The agent needs its own account on every service that you use, and in many cases needs its own credit card for each service. Agents are still largely experimental; there isn't a lot of structure given to them when they're granted access to accounts with a payment system. This can easily lead to runway spending, paid services getting silently canceled, or payments for products that are not needed. Incorrect decisions by the LLM can lead to real loss of funds, which is unacceptable, yet unavoidable given the current state of agentic AI.

In my own experience, I've only found some services like Cloudflare, DynaDot, and a few others with granular access controls. This feature restricts agents to using actions from a list they're constrained to, and as a result they don't have full control over your account.

You can go one level above this access control list and spin up a new Privacy Dot Com debit card with daily/monthly limits and a spending cap, or in some cases you can preload cash into your account (depending on the service, cloudflare for example does not have this). In most cases there needs to be an attached credit card. Many times the temporary or restricted cards get declined or are incompatible with the service due to payment gateways, regional restrictions, or another unknown reason due to tradfi being old and stupid. True story: my Venmo account stopped working because it says my "ZIP Code is invalid" (???).

TL;DR, most websites and service providers online haven't built their payment systems and payment rails in a way that's agent friendly, and AI agents still don't come out-of-the-box with a reasonably secure structure to avoid making costly mistakes.

If you yololog your agent into an Amazon account with no guardrails then you will inevitably find that it made the wrong decision or was able to do too much with the access that it was given in an insecure and unrestricted way, leading to financial loss.

It's still early, and one thing is abundantly clear to me, agents are the future for a lot of tasks. They just need to grow past this experimental stage into usable products for normal people who don't have finance or tech expertise.

[philip @philipglazman]

agentic payments

[psykeeper 𐁉 @psykeeper]

Zillions or Zero?

The most shocking thing about the zcash:native exploit is the human element. It took a person to prompt "hey check that codebase rq" for it to be exploited.

At this stage of the game I would have expected an agent to find something like this weeks before a human.

[psykeeper 𐁉 @psykeeper]

Current state of the market 🩸

[psykeeper 𐁉 @psykeeper]

@centrifuge @PaulFrambot @Morpho @rwasummit Fixed yield season

[psykeeper 𐁉 @psykeeper]

Based

[psykeeper 𐁉 @psykeeper]

He cooked

[Arthur MacWaters @ArthurMacwaters]

it blows my mind that Dario is about to have a trillion dollar IPO after being a 3-star michelin chef

insanely hard to do both

[pashov @pashov]

🤯An AI security tool has 1st-place performance on security contests from just 1yr ago. Solidity-auditor v3 is out, FREE & Open Source.

Thousands of Solidity developers are using the tool already. Upgrade your security baseline, use the tool🫡
https://t.co/SfxjuQ17gA

[psykeeper 𐁉 @psykeeper]

My openclaw lives on its own VPS, which means it can do stuff using installed programs that Codex can't do unless it's got access to install programs on your computer, which is a security concern.

It also creates a new website for each thing it does. My agent has been building an organized site with new pages for all of the knowledge it has accumulated over the weeks I've used it, including a few apps it's one-shot. People have found these apps useful for solving niche problems, and I can easily share hours of agent research with my peers to review it. This isn't really possible with an LLM harness alone, you need always-on filesystem access, and a webserver.

[psykeeper 𐁉 @psykeeper]

@marvv_eth We're still early™

[psykeeper 𐁉 @psykeeper]

@marvv_eth Yes. Reminds me of the Gnosis ICO and how "prediction markets" were like 10 years too early.

Also on the economic side: we're also still discovering how much tokens should *really* cost and what works and doesn't work as a business given the cost of compute

[Haseeb ＞|＜ @hosseeb]

Looking at on-chain revenue over the last 5 years, you can't really see cycles after 2022. Surprisingly consistent growth.

[psykeeper 𐁉 @psykeeper]

@NFTFlow_ x402 is the way, doing a research spike on that this weekend. From my preliminary research, the libraries that exist leave a lot to be desired, and X is actually a leader here (not surprising given the fintech origin)

[psykeeper 𐁉 @psykeeper]

Yeah

Exchanges

Users that trade on perps exchanges are simply losing money to exchanges based on path-dependency risk they are trading for zero upside

The slider is easier to understand than the strike price

[wale.moca 🐳 @waleswoosh]

Is anyone actually making money on perps?

$1,760,000,000 USD in liquidations