See you at SomethingTech

Over the past few weeks we’ve been in workshops, warehouses and cafés listening to local teams talk about what’s working (and what’s not) with AI. It’s reminded us that real progress is rarely flashy: it’s careful privacy choices, tidy workflows, and tools that actually get used. Next week we’ll be at SomethingTech with a small exhibition stall, showing our privacy-first local AI machines and how Gemini + NotebookLM can act as a simple “digital brain” for your business. If you’re around, come say g’day — even if it’s just for a quick coffee and a chat about where AI can give you back a few hours each week.

We’re exhibiting at SomethingTech (Brisbane)

We’ll be on the floor at SomethingTech, Thu–Fri 28–29 Aug, Howard Smith Wharves. Theme: Reimagining Human Intelligence — two days of Tech Talks, panels, trade show, River Rival pitch event, and new networks. If you’ll be there, swing past our stall to talk privacy-first AI and practical workflows with ChatGPT / Gemini / NotebookLM. (Advance Queensland, somethingtech.com)

Want to meet? Reply “Meet” with your preferred day/time.

News this week (relevant + immediately useful)

• Docs → Audio (read-aloud) in Google Docs. Gemini now generates natural-voice audio from a doc. Great for proofing on the go and accessibility. (Desktop, English; Workspace tiers + AI Pro/Ultra.) (The Verge, TechRadar)
• Gemini “August Drop”. New Guided Learning, Storybook (illustrated stories from prompts), Deep Think (Ultra), Temporary Chats (no history). Practical for training, proposals, and confidential chats. (blog.google)
• Gemini Live upgrades. Real-time visual guidance via camera plus smarter speech; useful for site walk-throughs / QA. (thehansindia.com)
• Pixel 10 AI features (shipping next week). Magic Cue offers cross-app suggestions; Gemini Live camera help; improved Voice Translate with voice-matching. Handy for field teams and travel. (MacRumors, Android Central)
• NotebookLM rollout progress. Video Overviews + Studio redesign now widely available; faster to turn a brief into narrated slides. (9to5Google, blog.google)

Feature: The “AI psychosis” headline cycle — signal vs noise

You’ve seen the “AI psychosis” chatter. Here’s the sober read:

• Anecdotes exist (some clinicians report cases); these often involve heavy usage plus vulnerabilities (sleep, isolation, pre-existing risks). (Wccftech, 4me.news)
• Evidence of causality is absent. Current briefings say there’s no peer-reviewed evidence that ordinary chatbot use alone induces psychosis; context matters. (aicommission.org)
• Media cycle ≠ mechanism. Even policy voices have framed this as bordering on “moral panic.” Sensible guardrails beat fear. (NDTV Profit)

Our stance: Using AI for research and drafting is net positive when done properly (clear questions, time-boxed sessions, triangulated sources, confidentiality controls).

Micro-playbook: Deep Work with AI (5 rules)

1. Write the exact question you’re answering.
2. 25-min research sprints → 5-min _model-off_ synthesis.
3. Ask for 3 sources per claim; verify in NotebookLM. (blog.google)
4. Prompt for red-team: _“List 3 reasons my conclusion might be wrong; cite.”_
5. For sensitive work, use on-prem or Temporary Chats. (blog.google)

Privacy-First AI Stack (SMB)

Deep Dive: Privacy-first AI you control (and how Gemini/NotebookLM fit)

Plenty of firms want AI without shipping sensitive data to remote servers. That’s why we’ve been building local AI machines (on-prem “AI appliances”) and pairing them with Gemini + NotebookLM.

Why local/on-prem?

• Data locality (control + risk reduction)
• Low-latency vision/QA on site
• Offline resilience

(Edge/on-device is broadly cited for privacy/security benefits.) (thehansindia.com)

Where Gemini/NotebookLM slot in:

• NotebookLM: curated “digital brain”, Video Overviews for explainers. (9to5Google)
• Gemini: reasoning, Storybook for training assets, Temporary Chats for confidentiality. (blog.google)

Quick comparator

| Use case | Local/Edge (on-prem) | Cloud (ChatGPT, Gemini etc.) | | --- | --- | --- | | Confidential docs | Best | Good w/ strict policies | | Vision on factory/site | Best (latency) | OK if bandwidth solid | | Heavy reasoning/creative | Optional | Best (latest models) | | Setup effort | Higher | Lower |

The stat that matters (fixed)

AI has moved from experimentation to execution. McKinsey’s latest survey: >75% of organisations use AI in ≥1 function, and ~21% of gen-AI users have redesigned workflows to capture value. Translation: the winners are rewiring processes, not just trialling tools. (McKinsey & Company)

Bonus stat — practical: Docs → Audio is already rolling this week across Workspace tiers and AI Pro/Ultra; desktop, English-only for now. (The Verge)

Where to find us next week

SomethingTech, 28–29 Aug (HSW). Come see how NotebookLM becomes your team’s “digital brain” and other local AI demos we will have on display (somethingtech.com)

Can’t make it? Reply “Audit” for the 1-pager on our $2,500 AI Opportunity Audit.

Sources (key items)

• SomethingTech dates/theme/location. (Advance Queensland, somethingtech.com)
• Docs → Audio feature. (The Verge, TechRadar)
• Gemini “August Drop.” (blog.google)
• Gemini Live upgrades. (thehansindia.com)
• Pixel 10 AI features. (MacRumors, Android Central)
• NotebookLM updates. (9to5Google, blog.google)
• McKinsey adoption + workflow redesign. (McKinsey & Company)
• AI-psychosis discourse (risk for vulnerable users; “moral panic” framing). (Wccftech, 4me.news, aicommission.org, NDTV Profit)