LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
Four myths about staking $LINK.
Anyone can deposit to get staked, no reSDL needed. Rewards grow your staked $LINK balance automatically with no further action to take.
stake.link stakes in @chainlink's own staking pools.
Staked $SDL earns a share of protocol fees in staked LINK - $stLINK.
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
JUST IN: 🇺🇸 JPMorgan CEO Jamie Dimon says the US economy is "booming."
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
Watching this 1960s dating show has completely fried my brain, what an era..😍😍
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
Security. Reliability. Institutional-grade compliance.
These are the non-negotiable requirements for moving trillions in tokenized assets across blockchains.
@SergeyNazarov explains why CCIP 2.0 became the interoperability standard for global finance.
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
How much better can an agent get without retraining the model? Google Research open-sourced Harness², a framework that keeps model weights fixed and improves everything around them: prompts, guardrails, memory, tool guidance, subagents, skills, scripts and plugins.
For each task it proposes edits, probes them on that same task, and composes a refined harness from how the runs differed, then carries the lessons forward to future tasks. The authors report gains of up to 11.5 points on WorkBuddy-Bench and 4.5 points on JobBench for GLM-5 and Gemini 3.5 Flash. It's Apache 2.0 research code, and there's a local demo that needs no model credentials.
If you build agents, run that demo first, then read the lessons it saves and ask whether you'd have made the same edits by hand. 🔧
#AIagents
Link in Comments
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
JUST IN: 🇺🇸 Vivek Ramaswamy's 'Strive' buys 1,223 Bitcoin worth $100 million.
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
WALLET ACCUMULATES $3.01M $LINK FROM BINANCE IN 4 DAYS
A wallet received another 67.56K $LINK ($808.35K) from #Binance.
· Past 24 hours: 182.16K $LINK ($2.18M)
· Past 4 days: 246.8K $LINK ($3.01M)
Wallet: 0xB52DD10622a706355881b39ab399bfea90Fa5a0D
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
IT’S HAPPENING.
France is watching its streets descend into chaos.
Riots. Looting. Burning barricades. Police struggling to restore order.
Meanwhile, French 10-year bond yields are approaching 5%.
Decades of political denial, failed policies and reckless spending are catching up.
You can pretend your country is thriving while its streets burn and its debt spirals out of control.
But financial markets don’t care about political fairy tales.
BOND MARKETS NEVER LIE.
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
AI agents that hunt security bugs spend 60.4% of their tokens understanding code and reasoning about how to trigger the bug, so that is where cost cuts should aim.
AI agents hunting security bugs can burn millions of tokens without producing an input that crashes the program. Across 200 runs from 4 agents, 60.4% of tokens went to understanding code and reasoning about how to trigger the bug.
Existing cost-saving methods rarely helped. Only 24.4% of head-to-head comparisons with the plain agent kept success at lower cost.
So they built AVRI, a tool that returns related code together and saves the links between program input and the unsafe operation. On 20 new tasks, it cut total cost by 18.0% for Codex and 23.7% for OpenCode without lowering success or recall.
Before adding generic cost tricks to a security agent, measure where its tokens go and let it reuse what it learned about the code.
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
Imagine this. Your boss calls and asks for your help. It’s urgent. It’s definitely him: you can see and hear him clearly.
But what if it isn’t? Deepfake technology has evolved rapidly. AI can now replicate real faces, voices and behaviours using just a few photos, videos or audio clips found online.
The threat extends beyond the workplace; scammers can impersonate friends and family, preying on emotion and creating a false sense of urgency to make you act quickly.
So what should you do?
🔵 Don’t rush. Even if you believe you’re speaking to someone you trust, take a moment to assess the situation.
🔵 Verify. Hang up and call the person back on their usual number.
🔵 Use a safeword. Establish a secret phrase with close contacts.
🔵 Secure your accounts. Turn on two-factor authentication everywhere. Activate multi-step verification for large business transactions.
🔵 Report it. If you suspect a scam, contact local law enforcement immediately.
🔵 Train your teams. Business owners: make sure your staff are trained to recognise phishing, vishing and other social engineering tactics.
Wait. Validate. #DontTakeTheBait
Find out more ➡️ 🔗 interpol.int/en/News-and-Eve…
Funded by 🇯🇵
🤖 Made with AI
LinkTOAD goldnutter 𓆏⬡⬢🌊 𓀡 logic/acc truth/acc retweeted
GitHub Decoded: the follow up
Since yesterday’s keeta.md post, more has landed on Keeta’s public GitHub. Only the new bits:
1) Bivo’s digital dollar is getting a new name: $KBUSD (and $KBEUR for euros). A converter already live on Keeta swaps them 1:1 with Keeta’s own dollar, $KUSD. One Keeta dollar, many issuers plugging in.
2) HopNow is now in the code, paying out to US bank accounts by ACH and instant RTP payments. Testing for now.
3) Keeta can now pay people on far more blockchains: Arbitrum, BNB Chain, Polygon, Avalanche, HyperEVM and Sonic, on top of Base, Ethereum and Solana. Robinhood Chain is set up too.
4) Usernames are coming. Both Keeta networks can now look people up by username instead of a long address.
5) AI agents are now tested daily with real $KTA on the live network. Not a demo.
6) The repo behind Keeta Compute was renamed this morning from “agent‑skills” to “keeta‑inference”. It’s no longer a side project.
Connecting the dots (my speculation): more issuers, more banks, more chains, all settling through one Keeta dollar.
All public GitHub. Not financial advice.
@schenkty @KeetaNetwork $KTA
Keeta just quietly shipped something big on GitHub.
A brand new repo went live yesterday: KeetaNetwork/keeta.md
In plain English, it’s an instruction manual written for AI agents. Point any AI at it and it learns how to move real money on Keeta.
What those agents can do:
- Open a US bank account in your own name (Bivo)
- Pay out to banks in 60+ countries
- Send money to debit cards with Visa Direct
- Move funds to Base, Ethereum and Solana (Bridge, LayerZero)
- Verify identity once with Footprint
- Pay per API call with x402
Put that next to Keeta Compute, the AI marketplace I broke down yesterday (post quoted below), and the picture gets clear. Agents can earn, spend, bank and get paid out, all settling on one network.
@schenkty and @KeetaNetwork are building the payment rails for the agent economy, and most people haven’t clocked it yet.
keeta:native $kta