Hello there,
Welcome to AI Unplugged. Here are today’s updates:
OpenAI accelerates its lightweight “Garlic” reasoning model
Gemini 3 Deep Think begins rollout for Ultra and enterprise users
Anthropic’s CEO flags bubble risks as AI spending outpaces infrastructure
Tools and a practical prompt to expose manual time drains and automate high-effort tasks fast ⬇️
NEWS ALERTS
OpenAI’s “Garlic” model signals just how aggressively the company is moving after Gemini 3’s benchmark leap. OpenAI is fast-tracking a new reasoning model internally called Garlic that’s already outperforming Gemini 3 and Anthropic’s Opus 4.5 on coding and logic — while staying small enough to deploy cheaply. The real story isn’t raw power, it’s architecture: Garlic fixes earlier pretraining gaps and compresses large-model depth into a much lighter footprint. If this lands anywhere near early next year, it points to smarter, lower-cost models arriving sooner than expected — and that directly pressures how companies think about inference costs, automation scale, and vendor choice. OpenAI
Gemini 3 Deep Think’s rollout by Google marks a clear push into serious reasoning territory. Deep Think is designed for advanced math, R&D-grade logic, and long multi-step scientific workflows, and it’s already leading several benchmarks. Google is rolling it out across AI Studio, Vertex AI, and enterprise products, not keeping it locked in demos. If your teams depend on tight reasoning — finance, engineering, analytics — this strengthens Google’s claim as a foundational model provider for 2025 systems, not just a productivity layer bolted onto search and docs. Google
Anthropic CEO Dario Amodei’s warning about an AI bubble cuts through the hype cycle. He’s openly calling out parts of the industry for overspending and racing to build data-center capacity that may not come online in time, creating a gap between model ambition and infrastructure reality. He positioned Anthropic’s slower, enterprise-first strategy as a deliberate counter to that risk. For companies betting long-term on AI, this isn’t commentary — it’s a reminder to prioritize vendors with realistic timelines, infrastructure discipline, and delivery credibility over flashy roadmaps. Anthropic Dario Amodei
PRODUCTIVITY TOOLS
🤖 Hexowatch is an AI-powered monitoring platform that automatically tracks website changes so teams stop checking manually and start acting faster, freeing up time for actual work instead of busywork.
✍️ CreativAI is an AI writing tool built to turn rough ideas into clean, publish-ready content in a fraction of the time, cutting production cycles by up to 10× without sacrificing quality.
🏥 Merative applies AI to healthcare analytics, helping organizations make better clinical and operational decisions using trusted data and domain-expert insights, not black-box guesses.
📈 Trolly AI is an SEO-focused writing assistant that helps teams produce search-optimized articles roughly twice as fast by baking ranking logic directly into the writing workflow.
AI MARKET
💰 Funding Rounds
Protego Biopharma secured $130M to accelerate its AI-powered drug discovery engine, pushing faster therapeutic development from early research through clinical-stage pipelines.
Mixx Technologies raised $33M to scale its AI platform for molecular generation, aiming to speed up biotech innovation and reduce the trial-and-error drag in early-stage research.
💼 AI Roles
Cloud AI Strategic Negotiator role at Cloudflare (US)
Staff Machine Learning Engineer, Communication and Connectivity at Airbnb (US)
PROMPT GUIDE
AI Workflow Time Leak Detector
Purpose: You want a fast way to expose where your team is wasting manual effort and which recurring tasks should be automated immediately.
Prompt:
Act as an operations optimization consultant.
I will describe my team’s daily or weekly activities.
Your job is to identify:
The tasks consuming the most time or effort
Which of those tasks can be automated or simplified using AI or software
The most practical tools or methods to use (for example: ChatGPT, Zapier, Sheets AI, CRM automations)
A realistic estimate of weekly time saved
Output format (table only):
| Task | Automation Recommendation | Tool / Method | Estimated Weekly Time Saved |Until next time,
AI Unplugged