Davie

AI Week in Review: July 28th, 2025 The AI News that I found interesting in less than 6 minutes

Yahoo Japan made GenAI mandatory for all 11 k employees to hit 2× productivity by 2028. That’s definitely one way to fast track adoption.

#AI #AIWeekInReview

open.substack.com/pub/dsweis/p...

AI Week in Review: July 21st, 2025 The AI News that I found interesting in less than 10 minutes

AI coding tools made experienced developers 19% SLOWER on complex tasks.
They expected to be 24% faster.

Sometimes the best ROI insight is knowing where NOT to deploy AI.

My weekend AI digest is live →
dsweis.substack.com/p/ai-week-in...

#AI #AIWeekInReview

Microsoft Says Its New AI System Diagnosed Patients 4 Times More Accurately Than Human Doctors The tech giant poached several top Google researchers to help build a powerful AI tool that can diagnose patients and potentially cut health care costs.

A very interesting question whether we are ready to trust AI with our lives. Microsoft claims 80% diagnostic accuracy with their new medical AI. Is it ready to diagnose your next illness? Not for 5+ years.

www.wired.com/story/micros...
#AI #HealthTech

Worried About Your Data? I Tried Running My Own LLM Privacy, control, and the future of edge AI, my experiment with local LLMs

Last week I hesitated before pasting my personal data into ChatGPT. Again.

So I tried running AI locally on my 4-year-old MacBook.

What worked, what didn't, and why ChatGPT admitted the local model wrote better code: open.substack.com/pub/dsweis/p...

#LocalAI #DataPrivacy #AI #LLM

Cloudflare launches a marketplace that lets websites charge AI bots for scraping | TechCrunch Cloudflare is launching a new marketplace that reimagines the relationship between publishers and AI companies.

Instead of blocking AI crawlers, publishers can now meter and monetize them. Cloudflare just created a toll booth for AI training data. Will AI firms pay up or find work-arounds? techcrunch.com/2025/07/01/c...
#AI #TechStrategy

US Senate removes controversial 'AI moratorium' from budget bill | TechCrunch After going back and forth over the provision, Sen. Marsha Blackburn offered an amendment to strip the provision alongside Sen. Maria Cantwell.

99-1 vote: Senate kills national ban on state AI laws. Keep on eye on this if you're building out AI systems. Could become very relevant very fast. techcrunch.com/2025/07/01/u...
#AI #AICompliance

Anthropic just won a landmark fair use ruling on AI training data.

Still need guardrails, but foundation model stability just improved significantly.

www.reuters.com/legal/litiga...

#EnterpriseAI #DigitalTransformation #LegalTech

Google provides cloud infrastructure to OpenAI while competing with ChatGPT.

The lines between competitor and partner continue to blur in enterprise tech.

Full analysis on Substack: dsweis.substack.com/p/strategic-...

Reimagining Global Business Services: Harnessing AI Agents The next productivity wave is building momentum, so make sure your people are riding it, not watching from shore.

AI agents = power steering for GBS.

Gartner: <1% of enterprise apps are agentic today—but by 2028 it’ll be 33%.

Pick 1 tough process, run a 6-week pilot, measure KPIs.
Next productivity wave is here—drive it.

#GenAI #GBS #automation

www.forbes.com/councils/for...

AI's arrival at work reshaping employers' hunt for talent An International Labour Organisation (ILO) report published in May found that around "one in four workers across the world are in an occupation with some degree of exposure" to generative AI models'…

Employers say the skills AI can’t replace: ethics, customer care, team leadership, strategy. Soft > hard in the AI era. #AI #HR #Skills economictimes.indiatimes.com/tech/artific...

AI's arrival at work reshaping employers' hunt for talent An International Labour Organisation (ILO) report published in May found that around "one in four workers across the world are in an occupation with some degree of exposure" to generative AI models'…

ILO: 25 % of all jobs touch GenAI, but “very few” risk total automation. It’s a reskill moment, not a layoff wave. #AI #FutureOfWork economictimes.indiatimes.com/tech/artific...

Germany—birthplace of Industry 4.0—just became home of Huang's bet on what that revolution really means.

The future of European manufacturing just got an address.

Full take: dsweis.substack.com/p/2948aa38-3...

#IndustrialAI #Manufacturing #Industry40

Why this matters: The data sovereignty angle: "Your data belongs to you. It is the history of your people, the knowledge of your people, the culture of your people."

European manufacturers can finally train proprietary models on proprietary data. In Europe.

Europe's going 10x on AI compute in 2 years. Not incremental. TEN TIMES.

Huang says Europe is "going all-in on AI." Partners include Siemens ("supercharging the company"), BMW, Mercedes, SNCF.

Why Germany? It's where the manufacturing already is.

Remember all those breathless digital twin demos vendors were pushing in 2019? Most were glorified CAD viewers.

This is different. Virtual wind tunnels. Factory twins where robots can actually train. Physics so accurate Nvidia partnered with Disney Research + DeepMind.

Nvidia announced the world's first dedicated industrial AI cloud in Germany. 10,000 GPUs including DGX systems.

Jensen Huang: "These AI data centers are not data centers at all... They have one job and one job only: to produce intelligent tokens."

Revenue gen, not cost centers.

Nvidia announced plans to build an industrial AI cloud in Germany with 10,000 GPUs. Focus: manufacturing use cases like digital twins and robotics. Key point: data stays in Europe, addressing sovereignty concerns. Part of broader trend of regional AI infrastructure. aibusiness.com/automation/n...

A clear answer on what tokens actually measure in AI. Tokens aren't just characters or words. They're chunks of meaning that vary wildly by language and context.

This chart from Mary Meeker's 2025 AI Report (slide 135) nails it:

Most unexpected chart in Meeker's report:

Autonomous taxis in San Francisco went from 0% to 27% market share between Aug 2023 and April 2025. Already beating traditional rideshare in some areas.

AI isn't just chatbots. It's already changing how we move around cities.

From Meeker's report: NVIDIA's share of global data center spending jumped from 10% to 25% in just 2 years (2022-2024).

The infrastructure buildout for AI is absolutely massive. Big Six tech companies are literally rebuilding the internet's backbone for AI.

The talent shift is already happening:

USA AI job postings ↑ 448% since 2018
USA non-AI IT roles ↓ 9%

Universities have ~5 years to adapt. Companies are scrambling for AI talent. The message from the data is pretty clear: this transition will be fast.

The speed is genuinely unprecedented:

ChatGPT → 800M users: 17 months
Netflix → 100M users: 10+ years

Meeker used the word "unprecedented" 300 times in the report. When an analyst repeats something that much, pay attention.

This isn't a Silicon Valley story anymore:

- 90% of ChatGPT users are outside North America
- India leads usage at 13.5% (US only 8.9%)
- Unlike internet/mobile, AI hit the entire world simultaneously

The geography of tech power is shifting.

The cost paradox is wild:

Training models: $100M → $10B+ 📈
Using models: Down 99.7% in 2 years 📉

How do companies recoup billions in training costs when usage is basically free? Even Meeker says "the jury is still out"

First big takeaway: Everyone will use AI, but only a handful can afford to build it

Training costs are heading to $10B+ per model. The "Big Six" tech giants spent $212B on AI infrastructure in 2024.

The AI oligopoly is real.

Trends – Artificial Intelligence BOND, May 2025

Recently discovered Mary Meeker dropped a 340-page AI report and not enough people are talking about it

Still working through it but already found some wild insights. Thread incoming...

Anthropic triples revenue to $3B in 6 months. Secret sauce: enterprise code generation with Claude 4. Specialized AI + B2B focus = explosive growth. I love Claude Code and can highly recommend it! #AI #Claude #AIBusiness www.reuters.com/business/ant...

Gemini 2.5: Our most intelligent models are getting even better At I/O 2025, we shared updates to our Gemini 2.5 model series and Deep Think, an experimental enhanced reasoning mode for 2.5 Pro.

Google's Gemini 2.5 sets new enterprise AI standards with Deep Think reasoning, native Model Context Protocol support, and built-in cost controls. I'm a big believer in Computer Use. Looks like we got Project Mariner. #AI blog.google/technology/g...

CES 2025: AI Advancing at ‘Incredible Pace,’ NVIDIA CEO Says Jensen Huang unveils NVIDIA Cosmos, Blackwell RTX 50 Series GPUs and AI tools for PCs.

HBR: GenAI slashes market research costs by 80%. Synthetic data, digital twins, automated insights. Meta's already doing it. Your marketing department's about to get superpowers. #AI #MarketingAI hbr.org/2025/05/how-...

Emergence AI’s new system automatically creates AI agents rapidly in realtime based on the work at hand The new system is literally a no code, natural language, AI-powered multi-agent builder, and it works in real time.

In general AI maturity in the enterprise is low. Which is why what Emergence AI is doing is pretty crazy. AI agents that create other AI agents autonomously. No-code, real-time, self-improving. Enterprise automation just went exponential. #RecursiveAI #Automation