Videos

The NEAR Protocol Working Group met to discuss and vote on NEP-655, a proposal for universal accounts.

I built a small open-source harness for benchmarking models on NEAR AI Cloud. You give it a prompt set, pick one or more models, and it scores every answer against what you expected and labels refusals as complied, partial or refused.

The NEAR Hackathon runs Nov 1–14, 2026, with $250K in prizes paid in NEAR. There are 13 build tracks, and this video walks through each one: what it asks for, which docs to open first, and a starter idea you could build in two weeks.

NEAR AI runs some models inside TEEs and proxies others to providers like OpenAI, Anthropic and Gemini. Both work with the same API key, but they give you very different privacy guarantees. Proxied requests strip your identity and still leave your content readable by the provider. TEE-hosted models run inside an…

Time to test the NEAR Agent Market for real. I hired Launch Kit, a $5 marketing agent, to review my website and paid in USDC. While it waited to start, I hired a second marketing agent at about the same price. You’ll see every step from job details to escrow to tracking both jobs from the Hiring tab.

Specialized AI agents each do certain jobs well, and paying for every model and bot gets expensive fast. NEAR AI built the Agent Market so you can find the agent that fits a job and pay for that job alone.

Brave Leo can now run GLM 5.3 Flash on NEAR AI Cloud, inside hardware enclaves that keep your prompts sealed while the model runs. Every response comes back with an attestation you can check. In this video I explain how it works, then show you how to turn it on in about a minute.

What happens when a NEAR chunk producer goes offline, and how can the network recover faster?

MyNearWallet is being sunset. This walkthrough shows how to move your account and tokens to near.com using the built-in export tool.

MyNearWallet is being sunset at the end of 2026. If it's the only place your keys live, you need to move your account before then. This video walks through the built-in export tool, step by step, from MyNearWallet to Meteor Wallet, plus the other options if you'd rather go to near.com or export your private key…

Any API key you ship inside an iOS or Android app is effectively public. Anyone can pull it from the binary, and once it leaks you're stuck until you push an app update. This video walks through the architecture I use to call NEAR AI's private inference from a mobile app without exposing credentials or routing prompts…

NEAR AI is now a provider on OpenRouter. If you already route through OpenRouter, you can pin private inference for a model in a couple of clicks, with the queries decrypted only inside a TEE and no retention on the provider side.

A BAA does not stop anyone from reading your users' health data. It decides who is legally on the hook when something goes wrong. Private inference answers a different question: who can technically see the bytes while a prompt is processed. If you're building anything that touches PHI, you need to know which one you…

Every AI coding agent sends your code somewhere. This one sends it into a TEE.

A full build walkthrough. We're analyzing a bank statement with private inference on NEAR AI cloud, and proving afterward that nobody along the path could read it or alter the result.

Every AI app claims it protects your data. The claim lives or dies in the architecture, so this video is the code.

If you're already building on the OpenAI API and you've been wondering what it would take to move to private inference, the answer is smaller than you think. NEAR AI's cloud API uses the same setup OpenAI does, so migrating is a base URL change. One line, and you're running.

Last video I walked through triage, deciding where each request goes once you know what's in it. This one covers what happens before that, because a few things need to run before you make any AI call at all.

For the last decade, data security meant encrypting data at rest and in transit. AI breaks that model, because data cannot stay encrypted while a model is processing it. At some point it has to be readable, and that's the hole.

In the first video I said IronClaw is built so the model never handles your raw credentials. This one puts that to the test. I tried handing over my Gmail password to see what the agent would actually do with it.

If you're building a product on top of the OpenAI or Anthropic APIs, your users' data is going somewhere, and most of them have no idea where. That matters for any product. It matters a lot more when the product holds health information.

Here's a second IronClaw tip, and this one is fun. I set up a routine that pulls every Massachusetts pro sports score every morning, along with the outcomes and any related news like trades.

IronClaw is an open-source AI agent from NEAR AI that runs your tools without ever handing your credentials to the model. In this first walkthrough I set up the web interface, run a real research task, build a daily automation, and show where the privacy claims hold up versus where they depend on choices you make.

NEP 642: Account cost increase, presented by Jan (the author). We need the two reviewers (Jakob and Arseni) to say they did approve it. This NEP has already been approved asynch by two working group members (Marcelo and Michael), so no voting needed but this needs to be stated out loud