Five AI and tech shifts for builders

Reuters reporting through Techmeme says Intel has decided to use ASML's next-gen High NA EUV tools for parts of its flagship Panther Lake laptop lineup.

The report is explicitly tied to High NA tooling, which is positioned as key for Intel’s newer process goals.

Techmeme also relays the specific cost marker for one High NA machine: about $400 million.

That makes this a notable capex and supply decision for anyone tracking chip-ready software and performance baselines.

Epic Games and Google have just jointly withdrawn their attempt to retroactively settle the lawsuit that’s changing how Android app stores work in the United States — and that means Google will be forced to carry rival app stores inside of its own. In fact, Google tells the court, it’s ready to begin carrying third-party app stores on Wednesday, July 22nd.

In October 2024, Judge James Donato originally agreed that forcing Google to carry rival Android app stores within its own Google Play store for several years, and forcing it to share its own entire catalog of apps with those stores, would be the best way of undoing the company’s illegal monopoly over Android applications.

But Judge James Donato was skeptical he should abandon his original permanent injunction in favor of Google’s proposed “Registered App Stores” that users would have to sideload — instead of simply downloading third-party stores directly through Google Play. On Thursday, July 16th, both parties were set to appear in court to argue it again , but that may no longer be necessary.

Here’s Google’s full statement on withdrawing its proposed modifications to Judge Donato’s permanent injunction, via Google spokesperson Dan Jackson:

We’ve agreed with Epic to withdraw our motion to modify the US Court’s injunction rather than prolonging this process which creates uncertainty for the ecosystem. This allows us to focus on executing our recently announced global business model evolution to deliver greater app store choice, lower prices, and more opportunities for developers and users.

Miles Wang , an OpenAI researcher whose work includes using AI to accelerate scientific and biological discovery, is leaving the ChatGPT maker to launch a new startup focused on developing AI models for drug discovery, according to four people with knowledge of his plans. Several other OpenAI researchers are expected to join the new company.

Wang is in talks to raise about $200 million at a $2 billion valuation, two of the people said. Lightspeed is in discussions to lead the funding round, according to sources. Talks are ongoing, the deal may not be final and details could change.

Wang disputed the story’s funding figures and description of the company but did not specify the correct numbers or details. Lightspeed didn’t respond to a request for comment.

The funding discussions point to investor interest in applying AI to make breakthroughs in life sciences. Chai Discovery, a two-year-old startup developing AI models that can predict molecular interactions to identify new drugs, announced on Tuesday that it raised $400 million at a $3.8 billion valuation. (Co-founder Josh Meier also passed through OpenAI as a researcher.)

Wang’s new startup may be working on AI models that will help find new uses for existing drugs and possibly those that previously failed in trials, a couple of sources told TechCrunch. Finding new uses for FDA-approved drugs can result in significantly faster time to revenue than developing new drugs from scratch, as these medicines have already been tested for safety.

The HN entry points to the GitHub changelog announcement: Dependabot version updates now introduce a default package cooldown.

That means dependency update traffic and PR cadence can change even if your repository config stays the same.

The thread sits at 176 upvotes and 112 comments in the scrape snapshot.

For teams running strict dependency gates, this is a release-risk and release-velocity issue, not a cosmetic setting.

The thread title is clear: “How to stop Claude from saying load-bearing,” posted from the HN community with substantial engagement.

Its focus is not model release timing but control in everyday AI assistant behavior.

No raw transcript is provided in this scrape, only the linked post and activity counts.

The engagement suggests this is a practical reliability issue for teams building Claude-based workflows.