iOS 27 Isn't Just About Siri — A Wave of Practical AI Features
The major AI overhaul of Siri stole the spotlight at WWDC 2026, but according to TechCrunch, iOS 27 ships with a host of practical AI features beyond Siri. While Siri's evolution is the headline grabber, these understated yet handy features may be what people actually end up using day to day. See Apple's official announcement and the TechCrunch article for details.
Block Announces "Builderbot" — An AI-Native Developer Tool Suite
Block, the company known for Square and other products, has announced Builderbot, a new AI-powered developer tool suite. According to the company's blog, it aims to change the very way software development is done. How an AI-native toolchain gets integrated into development workflows — and where it goes from here — will be worth watching.
Running Qwen 3.6 on a Dual R9700 Setup — Benchmark Results Shared
Benchmark results from running Qwen 3.6 27B (Q8) on a dual Radeon R9700 setup are drawing attention in Reddit's LocalLLaMA community.
According to the poster, with MTP (Multi-Token Prediction) enabled on llama.cpp, they recorded decode speeds of 64–67 tokens/s at contexts of 10–13k tokens. Even at 125k context, throughput held at 45 tokens/s — practical performance even for long-form processing.
Another post shared results comparing ROCm, Vulkan, and vLLM on the same dual-R9700 setup. The vLLM + ROCm + AITER combination was the fastest, hitting 156 tokens/s on Qwen 3.6 35B-A3B and 69 tokens/s on the 27B model — comfortably beating llama.cpp (ROCm) at 106 tokens/s. For anyone running local LLMs on AMD hardware, this looks like valuable reference data.
Cloak — Letting AI Agents Use APIs Without Handing Over API Keys
Cloak, released on GitHub, is a tool that lets AI agents use external APIs without ever seeing the API keys directly. The security risk of handing credentials to AI agents has long been a standing concern, and an approach like this seems likely to see strong demand in real-world production scenarios. It appears to be an early-stage open source project at this point, but the idea itself is an intriguing one.
The "Next Bottleneck" the AI Boom Faces in the Labor Market
A FiveTakes article points to workers as the next challenge in AI adoption. Beyond advances in the technology itself, the process of frontline workers accepting and mastering AI tools has become a major barrier. It's a fresh reminder of how important it is to look beyond technology when thinking about deploying AI across society.
---