Iroh 1.0, Qualcomm Chip Deal, Dev AI Updates

We think this is how the internet should work, which is why iroh exists, and today we're delighted to announce iroh version 1.0.

This is our first stable release, but the project has grown significantly over the 65 versions that led to 1.0. iroh is already used all over the place. The public relays we run have seen more than 200 million endpoints created, in the last 30 days alone. Developers are using iroh to stream video, train large langua…

After more than 4 years of building in the open, we have a foundation we're both proud of.

The power of that key can't be overstated. We use it to secure the connection. And because all data that comes from the connection is secured by that key, we can build up from that same key into identity, permissions, and attribution. We can also use that same key as an address we can dial, no matter where it i…

Iroh connections are also far more efficient. It's normal to see 95% of data transferred in a connection pass directly between devices. Going direct means fewer hops through the cloud, which lowers your egress bill. It's also fewer hops through routers, which means the internet is more efficient overall.

Qualcomm is in advanced negotiations to acquire Tenstorrent, a Canadian startup that designs specialized chips for AI model inference, for between $8 billion and $10 billion. The news was first reported by The Information and widely covered across tech industry outlets.

Tenstorrent's chips are designed to be more energy-efficient and cost-effective than Nvidia's dominant GPUs for running trained AI models, making them a popular choice for companies looking to deploy AI at scale. The startup raised $800 million in funding last year at a valuation of roughly $3.2 billion.

If the deal closes, it would mark one of the largest AI chip acquisitions in history and give Qualcomm a major competitive edge in the AI inference market, which is projected to grow exponentially as AI adoption increases across industries.

Industry analysts note the deal is a strategic move by Qualcomm to reduce its reliance on the mobile chip market and capitalize on the AI boom, which has driven record demand for specialized AI hardware.

A popular Ask HN thread posted on June 15 has sparked a widespread debate among developers about the viability of using locally hosted open-source AI models for daily coding work, instead of paid cloud tools like Anthropic's Claude or OpenAI's GPT.

The thread has garnered 596 upvotes and over 300 comments, with users sharing detailed experiences of running models like Llama 3, Mistral Large, and Qwen 2.5 on local hardware via tools like Ollama, llama.cpp, and LocalAI.

Proponents of local models cite benefits including zero ongoing API costs, offline access for work in restricted environments, and full control over code and data sent to the model, eliminating privacy concerns associated with cloud AI providers.

Critics note that local models often underperform compared to the largest cloud models on complex coding tasks, require significant upfront hardware investment (typically 32GB+ of RAM or a high-end GPU), and require manual tuning and maintenance to run reliably.

Popular low-cost cloud provider Hetzner has announced a broad price adjustment for its cloud server and dedicated server offerings, with most tiers seeing increases of 10% to 30% depending on region and hardware specifications.

The changes, published in Hetzner's official documentation, are attributed to rising energy and infrastructure costs across the company's global data center network. The price hikes go into effect for all new customers immediately, and for existing customers at their next billing cycle.

The news sparked widespread discussion on Hacker News, with over 440 comments from developers and small business owners calculating the impact on their hosting budgets. Many users noted that Hetzner's pricing was already competitive, but the hikes may push some to evaluate alternative providers.

Hetzner has stated that it will continue to offer its entry-level cloud server tier for €4.15 per month, but higher-tier instances with more CPU, RAM, and storage will see the largest price increases.

Over time, AMD added TSME to lower-end processors, including the consumer version of its Ryzen chips, a CPU that costs less than the Pro version. Over the years, users of these lower-end chips have gotten used to the added security. Recently and without warning or notice, this lower-end line of AMD chips suddenly dr…

AMD has yet to say why TSME worked on these CPUs, or even to confirm the change. AMD declined to answer questions sent by email other than to say TSME “is a security feature only applied to PRO CPUs as part of AMD PRO Technologies.” The statement is the first known time the chipmaker has explicitly made this restric…

In April, Ben Kilpatrick, who describes himself as a “privacy-conscious Linux hobbyist,” was installing a new OS on his machine running a Ryzen 7 9700X from the Zen 5 architecture. To check that all security protections were enabled, he had his machine run Host Security ID (HSI), an auditing feature that evaluates t…

To his surprise, HSI showed TSME was no longer possible, as indicated by the “encrypted RAM: not supported” line near the bottom of the screenshot below. A few lines lower, the HSI indicates that previously, TSME had shown as “encrypted.” This made no sense to Kilpatrick because he had enabled TSME in his BIOS setti…

This sent Kilpatrick into a monthslong investigation to figure out what had happened. After sending an inquiry to both the support and engineering teams at MSI, the manufacturer of his motherboard, he finally convinced company engineers to run tests.