ULTRA Loading

Initializing System

Skip to content
← Back to blog
aitechnologyai-roundupanthropicopenai

AI & Tech News Roundup: June 2, 2026

PewDiePie drops Odysseus, a free self-hosted AI workspace, and tells big tech the war is on. Plus Anthropic nears a $900B valuation that would top OpenAI, Karpathy joins Anthropic, Claude Opus 4.8 ships, and OpenAI answers with an IPO filing and a cybersecurity play.

June 2, 20265 min readBy Ultra Labs
AI & Tech News Roundup: June 2, 2026

AI & Tech News Roundup: June 2, 2026

PewDiePie drops Odysseus, a free self-hosted AI workspace, and tells big tech the war is on. Anthropic closes in on a $900 billion valuation that would top OpenAI, Andrej Karpathy defects to Anthropic, Claude Opus 4.8 ships, and OpenAI answers with a confidential IPO filing and a cybersecurity play. Here is what mattered this week.

The AI race got a wildcard this week. The most-discussed launch did not come from a frontier lab in San Francisco but from a YouTuber telling big tech the fight is on. Meanwhile the labs kept swinging: a record valuation, a marquee defection, a new flagship model, and the cloud giants crashing the coding party. Here is what happened.

PewDiePie Ships Odysseus, a Self-Hosted AI Workspace

The most talked-about AI launch of the week did not come from a lab at all. It came from a YouTuber. On May 31, Felix Kjellberg, better known as PewDiePie, released Odysseus, a free, open-source, fully self-hosted AI workspace that runs language models on your own hardware. He capped a roughly year-long build, documented publicly on his channel, with a blunt message to the incumbents: "The war on big tech has just begun."

Odysseus is not a toy. It bundles multi-turn chat, autonomous agents, built-in tools with MCP support, an email assistant, multi-step deep research, side-by-side model comparison, persistent memory, and even self-evolving skills into one local-first interface. Its built-in "Cookbook" offers hardware-aware recommendations and one-click serving across more than 270 catalogued models, and it plugs into vLLM, llama.cpp, Ollama, OpenRouter, or a plain API if you want one. The pitch is privacy by default: local-first, no telemetry, open source on GitHub, and free to clone and run.

The reason this resonates here is the same reason self-custody resonates in Bitcoin. Owning your own keys, your own miner, or your own model is the same instinct: keep the infrastructure, and the data, in your hands instead of renting it from someone else's cloud.

Anthropic Is About to Become the Most Valuable AI Startup

Anthropic is closing a funding round of at least $30 billion at a valuation north of $900 billion, a deal that would vault it past OpenAI's $852 billion mark and make it the most valuable AI startup in the world. Sequoia Capital, Dragoneer, Altimeter, and Greenoaks are set to co-lead, each putting in roughly $2 billion, with existing backers including Founders Fund and General Catalyst expected to participate. The round is expected to close within weeks.

The number is staggering, but the logic is straightforward: enterprise demand for Claude has been compounding, and investors are paying up for the lab whose models are increasingly the default choice for serious software and agent work.

Andrej Karpathy Leaves OpenAI for Anthropic

In a move that landed hard across the industry, Andrej Karpathy, a founding member of OpenAI and one of the best-known researchers in the field, announced he is joining Anthropic. He has started on the pre-training team, the group responsible for the large training runs that give Claude its core capabilities.

Talent moves like this are a signal. When a founding figure from one lab joins a rival, it says something about where the most ambitious people think the frontier is heading.

Claude Opus 4.8 Ships

Anthropic released Claude Opus 4.8, an upgrade to the Opus 4.7 flagship that went generally available earlier in the cycle. The company points to gains in coding, agentic work, reasoning, and knowledge tasks, with particular improvement on long-running jobs that require the model to verify its own outputs before reporting back.

The cadence here matters as much as the capability. Anthropic is shipping flagship-level improvements on a tight schedule, and pairing them with the funding to keep the training runs coming.

OpenAI Answers: Confidential IPO and a Cybersecurity Play

OpenAI is not standing still. The company is preparing to file confidentially for an initial public offering, working with Goldman Sachs and Morgan Stanley, with a listing possible as soon as September 2026 and a private-market valuation around $730 billion. A public OpenAI would be one of the largest tech listings in history.

On the product side, OpenAI unveiled Daybreak, its cybersecurity vision built around Codex Security. It includes a new GPT-5.5 Trusted Access for Cyber tier alongside a GPT-5.5-Cyber model for authorized red teaming and penetration testing, automating threat modeling, vulnerability identification, and remediation. Security has quietly become one of the most competitive fronts in AI, and both major labs are now planting flags on it.

Google and Microsoft Crash the Coding Party

The frontier labs no longer have the field to themselves. Google and Microsoft both rolled out AI coding models this week aimed squarely at the territory Anthropic and OpenAI have dominated. Google's broader push, unveiled at its developer event, included a world model called Omni, a general-purpose agent called Gemini Spark that can reason across a user's connected apps, and Gemini 3.5 Flash, a lighter model that delivers cutting-edge capability at roughly a third to half the cost of comparable options.

The competitive picture is widening fast. Cheaper, faster models from the cloud giants put pricing pressure on everyone and accelerate the pace at which capable AI shows up inside the tools people already use.

Why This Matters for Crypto and Mining

The thread running through all of it is compute and power. Every one of these announcements depends on more GPUs running in more data centers drawing more electricity, and that demand is reshaping the infrastructure layer that Bitcoin miners helped pioneer. We dug into the convergence in when AI meets crypto, and into the energy side in AI is burning through power, and Bitcoin miners already know how to handle it. And as the Odysseus launch shows, the self-hosting instinct that powers Bitcoin self-custody is spreading to AI: as agents take on higher-stakes work, the privacy and verification problems get harder too, which is exactly the gap projects like Midnight are building for, as we covered in the KYC problem Midnight might actually solve.


Ultra Labs is a US Bitcoin mining and technology company powered by renewable energy and built on Cardano. Delegate to the ULTRA pool using Eternl or Lace, or learn more at ultra-labs.io.