Breakthroughs and Backlash
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Saturday, August 22, 2026
Cerebras just shipped GPT-5.6 Sol at 750 tokens per second -- fast enough to chew through 2,500 PhD-level exam questions in a single workday. Meanwhile, Anthropic's IPO roadshow is quietly flagging public backlash against AI as a material risk, and Hollywood writers are getting paid to train the models replacing them. This Saturday's news reads like a species trying to build a rocket while arguing about whether rockets are a good idea.
Today in AI:
- Your Next AI Answer Arrives Before You Finish Blinking - Cerebras and OpenAI launched Ultrafast Mode, running GPT-5.6 Sol at 750 output tokens per second -- 11x faster than Claude Fable 5 and 5x faster than Opus 4.8 Fast. Select API customers get access first, with wider rollout planned. Cerebras
- Anthropic's IPO Paperwork Will Say the Quiet Part Out Loud - Anthropic's upcoming prospectus will list public backlash against AI and data center opposition as key risk factors, per sources familiar with test-the-water meetings. A Gallup survey found seven in ten Americans oppose data center construction in their area. Anthropic
- Hollywood Creatives Are Training Their Own Replacements - Award-winning writers, directors, and producers are taking gig work paying $12 to $200 an hour to teach AI models screenwriting and production scheduling. One described it as being handed a shovel to dig the grave of their profession. theguardian.com
- Inner Mongolia Is Becoming Asia's AI Power Plant - Ulanqab, a city of 1.5 million, now hosts nearly 100 data centers with 12.5 gigawatts of committed capacity -- more than OpenAI's entire Stargate Project. DeepSeek, ByteDance, and Alibaba are building their own facilities instead of renting cloud compute. Soz6
- Apple Cuts 200 Jobs to Bet on Glasses Over Goggles - Apple laid off roughly 200 employees across Vision Pro and Siri teams, shifting resources from immersive video and legacy Siri toward Meta-style AI glasses and LLM-based Siri AI. The company says VisionOS isn't dead, but the next headset won't arrive until 2028 at the earliest. Apple
- AI's Three-Body Problem Gets a Name - Fortune's analysis frames the 2026 AI economy as three gravitational forces -- closed-source frontier labs, Chinese open-weight models, and application companies -- all pulling each other into unpredictable orbits. Palantir's Alex Karp warned enterprises are spending furiously on tokens with no matching productivity gain. fortune.com
- AWS Wants Agents Building Your Data Pipelines, Not Running Them - Amazon launched ADOP, a reference architecture where AI agents generate ETL code, quality checks, and compliance controls at build time, then hand off deterministic artifacts to production. The key design choice: no model calls in production unless you opt in. aws.amazon.com
- Voice AI Benchmarks Are Cheating, and Now There's Proof - Hugging Face researchers found that several top speech recognition models memorized benchmark transcripts so thoroughly they reproduced them even when the audio was silenced or contradicted the text. High scores masked systems that can't actually transcribe unfamiliar speech reliably. huggingface.co
Today's Takeaway:
Cerebras delivering GPT-5.6 Sol at 750 tokens per second while Anthropic's CFO fields investor questions about public hostility toward AI infrastructure captures a tension that pricing alone can't resolve. Speed and capability keep climbing -- Ultrafast answered 2,500 PhD-level questions in under 12 hours -- but the physical and social costs of sustaining that climb are now showing up in IPO risk disclosures and Gallup polls (Anthropic).
Ulanqab's 12.5-gigawatt data center boom (Soz6) shows Chinese firms solving the compute bottleneck by building in cold, cheap, coal-rich regions -- a strategy that will sharpen the backlash Anthropic is warning investors about. The uncomfortable truth for anyone building on these models: inference speed is now a commodity war, which means your competitive edge isn't which model you call -- it's whether your customers and neighbors still want you calling it at all.
Social license, not latency, is becoming the binding constraint.
🔧 Tool Spotlight
- Google Vids - Make a training or update video with an AI avatar of yourself, inside Google Workspace.
What it does: Vids builds a video storyboard from a prompt or from files already in your Drive. It can record your screen, generate an AI avatar from a selfie and voice sample, and produce narrated video with Veo. It edits collaboratively in the browser, the same way Google Docs does.
Who it's for: Anyone who records the same explanation over and over - onboarding, customer support walkthroughs, internal updates, project recaps.
Try this first: Take a process you have explained out loud more than twice this month and turn it into a 90-second Vid. Send the link next time instead of scheduling a call.
Free or paid: Included with Google Workspace Business and Enterprise plans, and with Google AI Pro and Ultra subscriptions. Full access needs one of those; there is no standalone free version.
The Bottom Line
The Pattern: The AI industry's acceleration and its friction are no longer happening in sequence -- they're simultaneous. The same week a company ships the fastest frontier inference ever, another company's IPO filing treats public opposition as a material financial risk. Speed and resistance are scaling together.
Our Call: Anthropic's S-1, when it drops, will dedicate more than two pages to community and regulatory opposition as risk factors -- more space than any prior tech IPO has given to public sentiment. More likely than not by October 31, 2026. We'll grade this one in a Friday digest.
Your Move: Search your county name plus "data center proposal" on your local government website this Saturday -- takes three minutes. If you already did this after a previous issue, check whether the project has entered a public comment or permitting phase. That status tells you whether the backlash Anthropic is flagging to Wall Street has reached your zip code yet.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
✨ How AI-Fluent Are You? - Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team