Microsoft confirms "dramatic speed" boost for searching "This PC" in File Explorer on Windows 11.
Windows design and research boss Marcus Ash confirmed that the speed and relevance improvements Microsoft is making to Windows Search are coming to File Explorer next, starting with “dramatic speed improvements” when searching across This PC.
That matters because This PC search is painfully slow right now.
I tested the same file in Windows Search and File Explorer. Windows Search found it in about a second.
File Explorer search across This PC took more than 8 minutes.
And this was not some overloaded machine with thousands of files. It was a clean-ish test PC with barely anything downloaded.
Windows Search has already become much better: cleaner UI, fewer distractions, local-first results, typo handling, two-character file search, substring matching, and a toggle to remove Bing and Store suggestions.
Now those file search improvements are finally heading to File Explorer...
In ~2 weeks, if everything goes well, these feature amendments will come to the $XRP Ledger for voting:
- Batch
- Confidential Transfers (privacy)
- Sponsored Fees and Reserves
- Permission Delegation
- Dynamic MPT
- A bundled fix
Performance improvements:
- Up to 40% memory reduction on some machines best case, at least 10-15% though.
- Reduced likelihood to get out of sync for nodes (and validators).
Lots of other less splashy improvements of the XRP Ledger code base ❤️
A wallet built in four tiers.
Every wallet has one secret that can move its funds. What matters is which parts can touch it. SciPHR runs each action through four tiers, and only one of them can sign.
The first tier is the security chip in your iPhone. It checks a live Face ID to gate access, but cannot hold the key that signs XRP Ledger transactions. That key lives one tier down, in the keychain, released for the instant of a signature and never leaving the device.
The third tier is the backend. It prepares an unsigned transaction and broadcasts the signed one your phone returns. It holds no key. The fourth is the public ledger, holding the account's identity, its recovery rules, and an encrypted backup that opens only on your device.
Think of a courier who carries sealed envelopes but holds no key to open them. Down the stack, the signing key sits in one place, the device in your hand.
Source: https://t.co/tsbl9jZ1aO
How can an LLM switch between low-, medium-, and high-effort reasoning? And how does an LLM learn to reason more or less?
I put together a “little” article explaining how these effort levels are implemented at inference time and during training.
An identity server that never holds a key.
AI agents are starting to act on blockchains, and acting means signing, which means holding a private key. We built xrpl-identity-mcp, an open-source server that lets an agent work with XRP Ledger identity without holding that key.
An MCP server is a tool an AI assistant can call. This one resolves DIDs, reads credentials, and prepares the transactions that change them, each returned unsigned. Signing happens in the user's wallet or the agent's signer, not in the server, which holds no seed and no signing path.
Before anything reaches the ledger, the server decodes the signed transaction and checks it against the intent it prepared. A signature that does not match what was asked for is caught before submission.
An agent can read and write on-chain identity while the authority to sign stays with the person. Custody and automation do not have to be the same thing.
Source: https://t.co/CdkSokRTl8
Key hygiene you can read off the ledger.
On a public ledger, every account's key setup is visible. Whether it still has its master key enabled, whether it assigned a separate signing key, whether it set up recovery signers, the whole configuration sits on-chain in the open.
We run a read-only crawler that reads the validated XRP Ledger and reports this posture on a public metrics page. It reads ledger state and holds no key, so the most it can do to an account is describe it.
Think of a building inspector who notes which doors are propped open while carrying no keys. Anyone can look up an account and read its posture straight from the ledger.
Because the ledger is transparent, how a network holds its keys is a measurable fact rather than a claim. A whole chain's custody posture can be surveyed without touching one key.
Source: https://t.co/foLUda8IZC
Claude Code artifacts can now call MCP connectors, letting you build dashboards and apps that can fetch information and take actions for each viewer on demand.
Available on Pro, Max, Team, and Enterprise plans. Not available on publicly-shared artifacts.
A wallet's on-chain list of trusted devices.
Using one self-custodial wallet across devices means copying the secret between them, and every copy is another place it can leak.
SciPHR keeps the key on each device and lists the devices instead. Every XRP Ledger account publishes an identity record, a DID document defined by the XLS-40 standard, holding a public key for each enrolled device and nothing that can spend. A new device generates its own key on the phone, and a DIDSet transaction signed by an authorized key adds it. Removing one works the same way.
Think of a directory of which badges open the door, where the badge itself never leaves your hand. The list is public, and only the account can edit it.
Because only an on-chain signature can change that list, a stolen phone or a phished Apple ID cannot add itself. An identity built from public keys lets one account span many devices with no shared secret to lose.
Source: https://t.co/E1NE26UhW9
A CLI means you can drop DoorDash into whatever you already run.
Wire it into your internal tools and office catering orders itself. Point an expense agent at it and DoorDash receipts pull themselves. Give an assistant your calendar and lunch shows up in the gap between meetings.
US + Canada limited beta. Waitlist below.
When starting a company, most people shy away from truly hard problems and go for base hits thinking it will be easier.
The opposite is true. Fewer people attempt the hard problems so there is less competition, and great people are more likely to join you on an impactful mission. Instead, start with something you feel in your bones must be possible or different in the world, even if it's not a business.
Then work backwards from there toward what steps will be required, and see how you can get your first dollar of revenue on an early step which enables the broader vision.
A wallet your phone generates on its own.
Most wallets arrive already made. A server or app generates the keys and passes them down, so something other than your device saw the secret when it was created.
SciPHR generates the XRP Ledger keys on your iPhone, stores them in the iOS Keychain, and releases them only behind a live Face ID. The backend receives no seed and holds no key that can sign. It prepares an unsigned transaction and broadcasts the one your device signs.
Picture a key cut inside your own house, where the locksmith mails you the blank and never sees the finished cut. If the maker never touched the working key, there is nothing on their side to steal or surrender.
A key generated on the device that keeps it makes custody something you can verify, not something you have to trust.
Source: https://t.co/tsbl9jZ1aO
Most of the agentic payments conversation is still about what agents can do. The harder problem is how they pay for it.
On XRPL, that's already solved: 3-5 second deterministic finality, no gas auctions, no ambiguous pending states. An agent doesn't need retry logic or polling loops, it just proceeds the moment a transaction confirms. That's the difference between infrastructure built for humans clicking "approve" and infrastructure built for machines making decisions in milliseconds.
We shipped the XRPL AI Starter Kit last month to make that real for developers: an MCP server for XRPL docs, an Agent Wallet Skill for Claude, and a tutorial that takes you from zero to a confirmed payment in minutes. With X402 support now live, agents can already pay for APIs and inference using XRP or RLUSD.
As a Premier Member of the x402 Foundation, excited to help make this kind of infrastructure the standard, not the exception.
We're introducing Claude for Teachers: free access to premium Claude capabilities for verified K-12 educators in the US, with a library of teaching skills and a direct connection to evidence-based curricula, mapped to academic standards in all 50 states.
https://t.co/5hZZijVPCV
Ripple is proud to join the x402 Foundation as a Premier Member.
As AI agents begin to take on more of the transaction lifecycle, they'll need a way to pay that's as fast and reliable as the way they already exchange data. We've been helping build that future on the XRP Ledger with the tools and infrastructure developers need to enable agentic payments, including support for x402 that enables agents to transact using $XRP and $RLUSD today.
We're looking forward to contributing to the Foundation's technical and governance work as we help advance open standards for the next generation of internet-native payments: https://t.co/9Bm1K4srUp
A recovery signer that cannot act alone.
If you lose the device that holds your keys, the account still has to recover without handing anyone custody. SciPHR does this with a signer list on the XRP Ledger, a set of accounts allowed to re-key it together.
Each signer carries a weight, and the account sets a quorum, the total weight a group must reach to act. Your other devices and a recovery-code key are signers. SciPHR is one too, but its weight sits below that quorum. Its signature alone never reaches the threshold, so it can help a recovery and can never move funds on its own.
Think of a document that needs two signatures, where SciPHR only ever holds one.
A signer kept below the quorum can assist a recovery without ever being able to move funds. The rule is enforced by the ledger, not by trust in SciPHR.
Source: https://t.co/tsbl9jZ1aO