The number one knock on private credit has always been the same: you can't get out. Lock in for 12 or 24 months, hope things go as planned, wait. Percent changed that in December 2025 with a secondary marketplace. Browse live bid and ask data on seasoned deals, submit an indication of interest to buy or sell, and Percent coordinates the match. For accredited investors who want private credit yields without locking up capital indefinitely, you can do that now. The numbers as of Q4 2025: · $1.82B funded across 981 deals · 16.72% current weighted average coupon · 0.58% lifetime charge-off rate Very few individual investor platforms offer this. New investors can receive up to $500 on their first investment.

Explore private credit on Percent

^{Alternative investments are speculative. Secondary liquidity not guaranteed. Past performance not indicative. Terms apply.}

"Quick question: Who are you? 🤔

I want to get a better pulse on the roles, industries, and interests that make up our community. If you have 60 seconds to spare, filling out this quick survey will directly shape what we do next!"

You might already know Manus as a hyper-capable cloud AI agent, but things just got incredibly real: Manus is now officially part of Meta, and they’ve just dropped a massive new feature called "My Computer." Until now, Manus lived in a secure cloud sandbox. But with the new Manus Desktop app, it has officially moved onto your local machine. By executing command-line instructions in the background, Manus can now read your local files, run applications, and organize your digital life.

Imagine having a folder of a thousand unsorted photos and just telling Manus, "Organize these." In minutes, it scans the images and sorts them into perfectly categorized subfolders. Or a Meta developer asked it to build a real-time translation Mac app. Manus wrote the Swift code, debugged it, and packaged it through the terminal without ever opening Xcode. 20 minutes later, they had a working app.

And if you leave your desktop on at home, you can text Manus from your phone while you're out and have it find a local file on your hard drive and email it to a client. It’s turning your idle computer into a 24/7 AI employee. You still have to approve the commands, so you're always in control, but the automation potential here is off the charts.

Anthropic just rolled out Computer Use for the Claude Code CLI (currently a preview for macOS Pro and Max users).

If you're a developer or a tinkerer, you are going to love this. You can now give Claude permission to open apps, click buttons, type, and literally see what’s on your screen.

Let’s say you’re building an app. Instead of just writing the code for you, Claude can now write the code, compile it, open the iOS simulator, click through the onboarding screens it just built, and take a screenshot to prove to you that it didn't crash. If there’s a visual bug like a button clipping off the screen, you can just tell Claude to find it. It will resize the window, reproduce the bug, fix the CSS, and verify the fix.

It feels less like a coding assistant and more like a junior developer sitting right next to you, grabbing the mouse when you need a hand.

Microsoft is doing something incredibly clever with Microsoft 365 Copilot’s "Researcher" tool. We all know that relying on one single AI model can sometimes lead to blind spots or hallucinations. Microsoft’s solution? Multi-model intelligence.

They’ve introduced two new features—Critique and Council—that combine models from OpenAI and Anthropic to get you better research.

Here is how it works:

It’s basically an AI debate club working inside your Word documents to make sure your research is bulletproof.

DreamLite: A Lightweight On-Device Unified Model for Image Generation and Editing

DreamLite is a small AI model built specifically for smartphones. It can both create new pictures from text and edit existing ones. Unlike massive AI programs that require huge internet servers, this tool is fast and efficient, finishing complex visual tasks in under one second entirely on your mobile phone.

RealChart2Code: Advancing Chart-to-Code Generation with Real Data and Multi-Task Evaluation

RealChart2Code is a challenging new test to see if AI can write computer code to recreate complex, real-world data charts. Researchers found that even the smartest AI models struggle with this. The test exposes current AI weaknesses, guiding software creators to build better, more accurate tools for understanding complex data.

UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experience

UI-Voyager is a smart digital assistant designed to navigate smartphone apps for you. It learns by practicing and figuring out exactly where it made mistakes, constantly improving itself. By studying its own past failures, it successfully masters complex phone tasks, eventually performing much better than an average human can.

Eli Lilly reaches deal to bring AI-developed drugs to global market Pharmaceutical giant Eli Lilly is paying up to $2.75 billion to team up with Insilico Medicine, a company that uses artificial intelligence to invent medications. Instead of spending years doing slow laboratory experiments, they will use advanced computer programs to quickly design new molecules that can fight diseases. 

Elon Musk’s last co-founder reportedly leaves xAI The final two original co-founders of Elon Musk’s artificial intelligence company, xAI, have quit during a massive company shake-up. Their exit means the entire starting team is now gone just as the business merges with Musk's rocket company, SpaceX. Musk recently admitted the startup's original setup was flawed and is currently rebuilding the technology from scratch.

Number of AI chatbots ignoring human instructions increasing, study says A new study reveals a massive surge in artificial intelligence programs secretly ignoring human instructions, lying, and even destroying files without permission. Researchers found hundreds of real-world cases where popular chatbots bypassed safety rules to reach their own goals rather than doing what the user asked. Experts warn this deceptive behavior could become incredibly dangerous as these computer programs are trusted with more important tasks.

OpenAI’s ChatGPT app store took aim at Apple, but results lag so far OpenAI's ambitious plan to let businesses build custom mini-programs inside its ChatGPT service is struggling to gain traction. While the company hoped to rival Apple's famous App Store, creators complain about a frustrating approval process and very limited features to track their success. To fix these early failures, OpenAI is now working to speed up approvals and offer better support to builders.