In partnership with
🌟 Editor's Note: Recapping the AI landscape from 01/13/26 - 01/19/26.
🎇✅ Welcoming Thoughts
Welcome to the 27th edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
Interviews are back; Good one today with a long-time PE Investor.
I built a mini rag tool using Claude Cowork while working on something else at the same time. Lives up to the name!
Microsoft released a guide to GEO for AI search visibility.
Very quiet week for the NVIDIA5.
As busy as last week was, this week was quite the opposite.
I still don’t understand why GPT removed voice mode on Mac.
OpenAI looks like they’re getting into BCI 🤯
Anthropic released their latest Economic Index report. Washington D.C., NY, and Massachusetts use Claude the most.
Wyoming does not use Claude.
I found an open source alternative to Claude Cowork for engineers out there.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: OpenAI // ChatGPT
GPT Wins the Week. In one of the slower news cycles we’ve had across the NVIDIA5, OpenAI did just enough to push themselves out in front, with a big assist by a new compute deal. Here’s what they did:
Cerebras Compute Deal: OpenAI signed a massive, multi-year $10B deal with Cerebras for up to 750 MW of wafer-scale systems focused on fast inference, a move to diversify beyond NVIDIA and lock in real capacity through 2028. Good move. More compute for OpenAI.
Ads in ChatGPT: OpenAI confirmed it will test ads in the U.S. for users on Free + Go plans, with ads clearly separated from answers. Ads are Excluded from sensitive topics, and are targeted off the current chat context. It was bound to happen eventually. Will be a huge revenue driver, but likely hurt user retention.
GPTGo Goes Global: OpenAI launched ChatGPT Go worldwide as a lower-cost on-ramp ($8/mo), giving users more usage than Free and access to GPT-5.2 Instant. Cheaper option than Pro for those who run into usage limits.
OpenAI is clearly gearing up for 2026 as a scale year. CFO Sarah Friar says revenue grew from $2B (2023) → $6B (2024) → $20B+ ARR (2025), and frames 2026 as the year of practical AI adoption. In other, futuristic news, OpenAI led a $252M seed round into Merge Labs, a non-invasive brain-computer interface (BCI) startup, signaling a bet on future human-AI interfaces. I believe in BCI for treating paralysis and other health issues. Not sold on it beyond that. Fascinating nonetheless.
From Top to Bottom: Open AI, Google Gemini, xAI, Meta AI, Anthropic, NVIDIA.
⬇️ The Rest of the Field
Who’s moving, who’s stalling, and who’s climbing: Ordered by production this week.
🟠 Anthropic // Claude
Claude CoWork for All: Anthropic’s “digital coworker” mode was rolled out to Pro ($20/mo) users, letting Claude work across your files and apps for real tasks. I’d recommend paying for Claude Pro if you don’t already: Opus 4.5 usage alone makes it worth it, and this adds on! More in PUC.
Sequoia to Fund Anthropic: Sequoia Capital, one of the worlds top VC’s, will participate in a big way in Anthropic’s next mega-round (rumored $10B+ at a ~$350B valuation). Sequia is also an investor in OpenAI and xAI. Good move by Sequia even with other investments. I’d invest in this round too.
New India lead: Anthropic hired former Microsoft exec Irina Ghose as Managing Director of India to lead its Bengaluru office, their 1st physical location in India. India is a massive market for AI. Important hire here.
🟣 Google // Gemini
Personal Intelligence Beta: Opt-in personalization that lets Gemini (with permission) pull context from Gmail, Photos, YouTube, Search, etc. to give tailored planning and recommendations. Cool but I don’t think I’ll opt in, AI doesn’t need to be that personalized for non-work data yet.
TranslateGemma Launch: Google released open translation models built on Gemma 3 covering 55 languages, tuned via distillation and reinforcement learning. Nice, essentially just improving their AI language models.
Higher Gemini Prompt Limits: Google raised daily caps for paid plans by separating limits for “Thinking” and “Pro” models, giving clearer control and higher daily usage.
🔴 xAI // Grok
Pentagon Deploys Grok: DoD confirmed Grok will run on Pentagon networks (classified + unclassified) for analysis and operations support. This works IF the right people are setting it up. xAi has already had trouble with leaks. Hopefully there is additional security parameters on classified material.
Musk v. OpenAI Advances: Musk’s lawsuit against OpenAI cleared the path toward a 2026 trial, the lawsuit is built around Musk’s early investment into OpenAI and its previous non-profit stature. This is gonna be a big and expensive trial. If I had to guess, I’d say Elon has the edge going into it.
EPA Tightens Data-Center Rules: EPA closed xAI’s “turbine loophole,” requiring permits for trailer-mounted gas turbines and potentially slowing the Memphis build-out.
🔵 Meta // Meta AI
Reality Labs Layoffs: Meta cut about ~10% of Reality Labs (reported ~1,000–1,500 roles) as it shifts resources fully towards AI. It’s clear: AI is the future at Meta, not VR.
Powell McCormick: “AI is a group sport”: Meta’s new president Dina Powell McCormick says AI needs rival cooperation on safety, energy, and “smart regulation” to keep humanity at the center. Sounds like she’s taking an approach similar to Anthropic. I like that.
Quest for Business Winding Down: Meta ends Quest for Business hardware sales Feb 20 and sunsets the platform Jan 4, 2030, another signal it’s prioritizing AI wearables over VR headsets. Another serious signal that Meta is moving away from the Metaverse.
⚪️ NVIDIA
New H200 Export Tariff: Following the U.S. recently reopening H200 chip exports, they added a 25% tariff/fee that effectively taxes those overseas sales. I don’t mind this. China is closing the gap fast in AI.
China Blocks H200s at Customs: Chinese customs reportedly stopped clearance for early H200 shipments, spooking suppliers and pausing some parts production. Back and forth we go. Chinese Govt wants companies to use in-country suppliers.
Eli Lilly AI Lab Blueprint: Following the partnership from last week, Jensen Huang and Lilly CEO Dave Ricks outlined details on how the 5-year $1B lab will work. Simulations → Testing → Trials. Very bullish on AI-based simulations.
🤖 Impact Industries 📢
Robotics // Human Interaction
Columbia Engineering researchers built a humanoid robot that learns realistic lip movements for speech and singing by observation, not rules. The robot first learned how its own flexible, motor-driven face moves by watching itself in a mirror, then studied hours of human videos to map sounds directly to lip motion. The result is more natural facial expression, helping robots cross the uncanny valley in social, educational, medical, and elder care settings.
Marketing // Avocados for AI
Instead of buying an $8M Super Bowl TV ad, Avocados From Mexico launched an AI-powered interactive experience featuring real-time football predictions and personalized guacamole recipes. The “Prediction Pit” uses a digital avatar of Rob Riggle and updates outcomes based on live variables like injuries and weather. With strict guardrails around content and likeness, the brand is betting AI-driven engagement beats traditional ads.
🎙 Weekly Interview: 10 Minutes With Ted Wong
Ted Wong
🏠 Background: Ted Wong is a Miami-based private equity investor with 15+ years of experience across credit, growth equity, and buyouts. He studied economics at Harvard and later earned a Wharton MBA (Finance) plus a JD from the University of Pennsylvania Law School. He’s also an active collector and investor in trading card games, especially Magic: The Gathering.
💼 Work: Ted is a Principal at The Firmament Group, where he helped build the firm from its founding and led its supply chain technology investment practice. He’s served on 15+ portfolio company boards and supported exits including Logistyx (E2open), Envase (WiseTech Global), and Principal Logistics Technologies (Aptean), while also contributing as a writer/moderator in the collectibles space.
🚀 Quote: “AI looks good and sounds good, but without critical thinking, you can’t tell whether what it’s giving you is real or not.”
🎙️ Condensed Interview Transcript — Ted Wong
1. Noah Weisblat:
What’s your high-level take on how AI has evolved over the past few years?
Ted Wong:
AI is game-changing, life-changing technology. Every business I’m involved with is either already using it or actively exploring how to improve internal operations and create more value for customers. It’s touching nearly every part of how companies operate.
2. Noah:
How do you use AI professionally in your investment workflow?
Ted:
I’m pretty old school when it comes to sourcing deals. I’m not using AI for mass outreach or drafting messages. Where it really helps is early-stage research and analysis—getting up to speed quickly, understanding competitors, and synthesizing information before deeper diligence.
3. Noah:
What advice would you give students entering college in the age of AI?
Ted:
College taught me how to think and analyze problems. That foundation is even more important now. AI gives you answers, but without critical thinking you can’t tell whether what it’s giving you is solid or nonsense.
4. Noah:
What excites or concerns you about the future of AI?
Ted:
A lot of jobs and roles will change, and some will become redundant. But I’m not bleak about the future. The real question is whether institutions adapt and teach people how to work with AI.
5. Noah:
Anything else you wanted to add?
Ted:
AI can’t replace human interaction. Community, hobbies, and real relationships matter. As AI becomes more dominant professionally, those human connections become even more important personally.
👨💻 Practical Use Case: Claude Cowork
Difficulty: Mid-Level
Claude Cowork is Anthropic’s newest step toward making AI agents more usable for non-developers. Think of it as a more guided, non-dev-friendly version of Claude Code. Instead of just chatting or writing code in isolation, Cowork is designed to help Claude actually work alongside you inside a project.
It’s only about a week old, so there are still some bugs. In my testing, it had trouble accessing files on my PC, which is one of the headline demo use cases. Still, the direction is clear, and it’s already useful.
Claude Cowork in Action
I used Cowork to set up a very basic RAG (Retrieval-Augmented Generation) tool. It was able to guide the setup end-to-end and write the scripts/configs I needed, but it couldn’t execute them on its own yet, so I ran them manually.
It helped configure:
Supabase: a hosted database, basically where the data lives
Vercel: a hosting platform, basically where the app runs
A few examples of what Cowork is built to help with:
Cleaning and organizing files: sorting docs into folders, renaming consistently, and generating a clean summary or action list.
Hooking up APIs and connectors: wiring together tools you already use (CRM, forms, email, Slack) so data flows automatically without copy-paste.
Building internal mini apps: simple internal tools like an intake form, lead tracker, meeting prep assistant, or a lightweight dashboard for your team.
Cowork feels similar to agent-style browsing (like what we showed with GPT Atlas), but it’s a bit safer and more constrained, especially when it comes to prompt injection. Overall, Cowork has received a ton of hype, and I think it’s going to live up to it over time.
Learn more below ⬇️
🩸 Startup Spotlight
Vitestro
Vitestro - Automated Blood Drawing for the Modern Clinic.
The Problem: Blood draws are one of the most common medical procedures, yet they still rely on manual labor, are prone to human error, and increasingly face staffing shortages. In high-throughput environments like hospitals and diagnostic labs, this bottleneck slows operations and strains resources.
The Solution: Vitestro has built an autonomous blood-drawing device that uses AI and ultrasound imaging to identify the optimal vein and robotically insert the needle: no human phlebotomist required. The device standardizes blood collection, reduces variability, and frees up healthcare staff for more complex tasks. It’s already being trialed in European hospitals and aims to become the go-to method for high-volume blood draws.
The Backstory: Founded in 2017 and headquartered in Utrecht, Netherlands, Vitestro combines robotics, AI, and medical imaging in a single device. The team includes engineers, healthcare experts, and former medtech execs, and the company has already raised multiple funding rounds to scale deployment. Their long-term goal is to make blood testing more efficient, consistent, and accessible; starting with automating the first step.
My Thoughts: Just inside the startup window, but a great use case demo. This is exactly where AI and automation should be starting out when it comes to ‘Physical AI’ in healthcare applications. Standardizing a heavily repeated process to free up time for healthcare professionals. The more operations like this that are automated, the more opportunity there is for innovation. Excellent.
“It’s not likely you’ll lose a job to AI. You’re going to lose the job to somebody who uses AI”
- Jensen Huang | NVIDIA CEO
Anybody in the market for a BCI implant? Till Next Time,
Noah on AI