It was a day of rapid-fire developments where AI intersects with employment, drug discovery, openness, privacy, and local communities. Since our recent published articles already covered the Opus 5 release and the open-weights regulation debate, this roundup focuses on topics that haven't been touched on yet.
monday.com Cites AI in Layoffs as 'AI Job Losses' Become a Catalogued Trend in Tech
monday.com has been added as the latest case to TechCrunch's running list of tech companies announcing major layoffs citing AI. The list tracks, in reverse chronological order, more than 20 major companies that named AI as an explicit factor during 2026 — and monday.com is its most recent entry.
What's notable is that companies have started putting 'AI-driven efficiency gains' front and center as an official justification for cuts. It suggests that 'AI-related layoffs' are becoming visible as a tech industry-wide trend rather than isolated incidents. That said, from the outside it's hard to tell what is genuinely 'because of AI' and what is simply a business decision, so it remains important to check each company's actual announcements rather than headlines alone.
The Wall in AI Drug Discovery: Design in Seconds, Efficacy Still Unknown
An essay published on philippdubach.com throws cold water on the AI drug discovery boom. Its argument: 'AI can now design drugs in seconds. But whether they work is still something no one can tell you.' It points out that, behind the race to optimize generation speed, validation remains a bottleneck in both time and cost.
AI has certainly accelerated molecular design and the narrowing of candidate compounds, but the wall of clinical trials — actually proving efficacy in humans — is not something AI can shorten. The observation that the gap between the time scales of 'design' and 'validation' shapes the reality of drug discovery reads as a healthy reality check on AI-omnipotence narratives.
Karpathy Removes Anthropic from His Bio; Speculation Links It to the Open-Weights Debate
As reported in Reddit's LocalLLaMA community, Andrej Karpathy appears to have removed the Anthropic mention from his X bio. He joined Anthropic just a few months ago, and the apparent departure after such a short stint has been met with surprise in the community.
The post speculates that it 'may be connected to Anthropic's strong opposition to open-weight and open-source models.' This is purely community speculation, however, and no official announcement has been confirmed at this point. Karpathy has long been known as an advocate of open-source AI and is also an OpenAI co-founder, so whether Anthropic's policies around openness and his personal stance played a role is a point worth watching.
Logue: A Fully On-Device, Privacy-First macOS Notes App Goes Open Source
Bitwize has released Logue, a macOS app for meeting notes and writing, under the MIT license. Its standout feature: both audio and text are processed entirely on the Mac, with no data ever sent to the cloud.
The tech stack centers on Swift + SwiftUI/AppKit, with MLX (mlx-swift-lm) running LLM inference, Apple's Speech framework handling real-time transcription, and FluidAudio providing speaker diarization. It is designed to keep everything local — meeting note summarization, action item extraction, 60+ writing modes, and even PII detection. Data is stored encrypted with AES-256-GCM.
On the other hand, it requires a fairly recent environment: macOS 26 (Tahoe) + Apple Silicon only. The project is in its early stages — the author themselves admits it has 'rough edges' — which also makes it valuable as reference code for developers who want to try implementing an MLX-based local AI pipeline.
The Hidden 'AI Tax' on Smartphones, Per an NYU Stern Expert
Fortune reports that Apple's AI feature requirements are imposing a 'hidden AI tax' on your next smartphone through rising memory and chip prices. The issue, flagged by a pricing expert at NYU Stern, paints a picture in which the hardware needed to support AI features is passed on to device prices in a form that is hard for consumers to see.
Better AI experiences require more memory and dedicated accelerators, and those costs are ultimately reflected in device prices. The perspective that AI may appear 'free to use' while you are actually prepaying for it at device purchase is a key talking point for thinking about the cost structure of AI adoption.
Data Centers and Local Communities: The Age of Towns Negotiating with Amazon and Microsoft
Also in Fortune: as AI data centers choose where to locate, local governments are increasingly negotiating 'community benefit' deals with operators such as Amazon and Microsoft. The featured example is the town of Hobart, illustrating a spreading situation in which residents proactively negotiate the balance between the burdens data centers bring to a region (power, water, traffic) and the returns (tax revenue, jobs, infrastructure investment).
It's an example of how AI infrastructure expansion now directly affects not just urban data centers but also living environments in rural areas. From the 'hidden costs' angle, it shares a theme with the smartphone AI tax above, and how to design the distribution of AI's benefits and burdens looks set to become a key debate going forward.
---