In partnership with
🌟 Editor's Note: Recapping the AI landscape from 01/20/26 - 01/26/26.
🎇✅ Welcoming Thoughts
Welcome to the 28th edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
Great Interview this week with someone in-tune to both AI and Quantum.
AI may soon be used in drafting our legislation.
Lots of AGI buzz right now → Deepmind is hiring a Chief AGI Economist.
Reports indicate the new OpenAI device could be earbuds.
“ClawdBot” is taking tech Twitter by storm, more in the Race and PUC.
Another somewhat quiet week amongst the NVIDIA5.
Anthropic CEO released a detailed essay outlining some warnings on risks once AGI arrives.
Shocking: xAI is still in legal hot water.
NVIDIA’s new voice model is crazy accurate.
Please warn the older adults in your life about AI spam calls. They may sound near indistinguishable in the coming years.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: Anthropic // Claude
Welcome Back Anthropic: Another quieter week across the AI field, but Anthropic still managed to go viral while progressing across workflows, education, and the broader agent ecosystem, reinforcing Claude’s positioning as the premier enterprise model.
Claude Apps is Here: Anthropic expanded MCP and launched Claude Apps, letting users use apps inside the Claude chat to take actions, not just summarize. Includes Slack, Asana, Figma, Canva. These releases aren’t perfect yet, but Claude is making huge long-term strides for enterprise AI.
Teach For All partnership: Anthropic partnered with Teach For All (63 countries) to help 100,000+ teachers/alumni build AI fluency and share classroom projects via Claude Connect. It’s framed as “teachers as co-creators,” not a top-down rollout. Good move.
Clawdbot Goes Viral: Clawdbot, an open-source personal assistant from PSPDFKit founder Peter Steinberger, is taking off with early adopters. It runs locally, and can do multi-step tasks with real system access. More on this in Practical Use Case.
Also worth noting: Anthropic added Mariano-Florentino Cuéllar to its Long-Term Benefit Trust, Dario Amodei blasted the H200 China approval with a “nuclear weapons to North Korea” comparison, and Anthropic published an updated Claude Constitution outlining the principles used to shape Claude’s behavior. The Claude Constitution almost seems like it was written for Claude… slightly dystopic.
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.
🟢 OpenAI // ChatGPT
ChatGPT Age Prediction: OpenAI is rolling out age-estimation to auto-apply stricter teen safety rules; misclassified adults can verify via Persona to restore access. Probably smart.
Gates Foundation Partnership: OpenAI and the Gates Foundation launched a $50M program to bring AI support to 1,000 primary-care clinics in sub-Saharan Africa by 2028, starting in Rwanda. I really like the application of AI in healthcare. This will be interesting to follow.
Stargate Community Plan: OpenAI says its data-center buildout will “pay its own way” on power so local rates don’t rise, funding grid upgrades/new supply plus local workforce benefits. Good move. Prepping for potential legislation pushing this.
🟣 Google // Gemini
Gemini in Education: Google added free SAT practice in Gemini and partnered with Khan Academy to power an AI Writing Coach that helps students outline, draft, and revise instead of writing essays for them. These would’ve been helpful when I was in high school!
No Gemini Ads: Google says Gemini is staying ad-free for now, with leadership publicly pushing back on ads inside the chatbot experience. Nice, interested to see if this changes as generic search and AI usage gets closer.
DeepMind Acquisitions: Google DeepMind did two small acqui-hires: a licensing + talent deal with Hume AI (to boost Gemini voice) and a pickup from Common Sense Machines (to speed up 2D→3D features). Big boost for ‘Physical AI’ and AI Voice features. Voice is improving industry-wide.
⚪️ NVIDIA
$2B More For CoreWeave: NVIDIA put another $2B into CoreWeave to deepen the partnership and help scale GPU-heavy data center capacity, with funding geared toward infrastructure buildout. Slight shift of these large investments going to energy/infrastructure vs. chips in the past few months.
New Voice Model: NVIDIA open-sourced PersonaPlex, a full-duplex voice model that can listen and talk at the same time for more natural, interruptible conversations (English-only). I listened to a PersonaPlex conversation and wow. AI voice is getting shockingly close.
New AI weather models: NVIDIA released three open weather models (15-day forecasting, 6-hour nowcasting, and data assimilation) aimed at making forecasts dramatically faster and cheaper to run at scale. Interesting insurance applications here.
🔵 Meta // Meta AI
Teen Access Paused: Meta is temporarily blocking teens from interacting with its AI “character” bots across its apps while it rebuilds the experience with stronger safeguards and parental controls. Take notes @xAI.
New Lab Ships First Models: Meta says their Superintelligence labs have produced and deployed the first high-profile internal models for evaluation, with codenames like Avocado (text) and Mango (image/video). This is what all the spending has been for… lets see if it lives up to internal hype.
Premium Meta AI: Meta will test new paid tiers on Instagram, Facebook, and WhatsApp that add enhanced AI features (including Manus and Vibes), while keeping core app access free. Intrigued as to what they do with Manus.
🔴 xAI // Grok
United States AG Pressure: Dozens of state AGs demanded xAI explain how Grok will stop non-consensual sexual deepfakes and remove existing harmful content. Good. Fix it.
National Probe: Brazil gave xAI 30 days to identify and remove harmful sexualized AI content and disable offending accounts. The EU opened a Digital Services Act investigation into whether X/Grok did proper risk mitigation before rollout.
Elon spoke at Davos: Musk framed AI + robots + cheap energy as a path to “abundance,” said energy (not chips) is the main constraint, floated space-based power/compute, and reiterated Optimus/robotaxi roadmaps. Elon is often optimistic in his timelines but generally has a good idea of what’s to come.
💻 Impact Industries 🚑
Software // Real-Time World Gen
Startup Overworld open-sourced Waypoint-1, a generative AI model that creates interactive 3D game environments live from text prompts and player inputs. Trained on 10,000 hours of gameplay, Waypoint-1 streams playable worlds at 30–60 FPS on consumer GPUs using its new WorldEngine. The release points toward faster game prototyping, AI-assisted world building, and entirely new forms of player-driven content creation.
Healthcare // FDA Approved AI
Aidoc received FDA clearance for CARE™, the first foundation model AI designed to triage multiple acute conditions in a single CT workflow. The system detects 14 abdominal emergencies at once, achieving 97% sensitivity and 98% specificity while reducing false alerts by roughly 10x versus prior tools. Already deployed across 1,600+ medical centers serving 60 million patients annually, the approval signals growing regulatory confidence in multi-purpose clinical AI.
🎙 Weekly Interview: 10 Minutes With Caden Kacmarynski
Caden Kacmarynski
🏠 Background: Caden Kacmarynski is a physicist and AI architect bridging the gap between quantum optics and machine learning. He holds a degree in Engineering Physics and is completing a Master’s in Entrepreneurial Physics from Case Western Reserve University. His technical foundation includes hands-on quantum research at Case Western and Argonne National Laboratory.
💼 Work: Caden serves as the Chief AI Architect at Moreland Connect, where he designs custom AI solutions for healthcare, manufacturing, and enterprise applications. He is also a Perplexity Business Fellow, the Chair of the Quantum Coalition, a nonprofit advancing quantum education, and the recent founder of startup Reborn Career AI. You can connect with Caden via email here.
🚀 Quote: “Software engineering AI is 2 years ahead of knowledge work AI because code is structured and business data is fragmented.”
🎙️ Condensed Interview Transcript — Caden Kacmarynski
Question 1
Noah Weisblat: Where do you see the current state of the AI space?
Caden Kacmarynski: I think the space is both over-hyped and undervalued at the same time. There’s a lot of noise, but the opportunity is bigger than the dot-com era because the infrastructure already exists. We don’t have to lay the lines ourselves. We just have to take advantage of what’s been built.
Question 2
Noah Weisblat: What does your personal AI tool stack look like?
Caden Kacmarynski: I lean heavily into the Anthropic ecosystem. I use Cursor and Claude Code daily as a developer. I rotate through 50 to 100 tools overall, but those two are essential. They let me stay productive and even “vibe code” entire websites for presentations.
Question 3
Noah Weisblat: How do you explain quantum computing to a non-technical audience?
Caden Kacmarynski: Think of a traditional computer like a light bulb and a quantum computer like a laser. A light bulb is great for everyday tasks. A laser is a high-power tool used for very specific, complex problems. We have light bulbs everywhere, but we don’t all need to carry a laser.
Question 4
Noah Weisblat: What advice would you give a high school or college senior entering this market?
Caden Kacmarynski: Become the best AI-enabled version of yourself. Understand the difference between “forklifting” your work and “weightlifting” your skills. Use AI to strengthen your thinking, not replace it. Solving hard, systems-level problems gives you the ability to supervise AI effectively.
Question 5
Noah Weisblat: Where do you see quantum computing making real-world impact?
Caden Kacmarynski: Energy is a major one. Quantum systems are far more energy-efficient for certain workloads. As AI pushes us into energy constraints, quantum could either help run those algorithms or simulate breakthroughs in areas like fusion and materials science.
👨💻 Practical Use Case: Clawdbot
Difficulty: Advanced
Clawdbot is an AI assistant designed to live where you already communicate, not in a separate chat window. Instead of opening ChatGPT or Claude, Clawdbot works inside apps like iMessage, WhatsApp, Telegram, Slack, or Discord. You message it like a person, and it responds in the same thread, with memory and context that carries over time.
Many people run Clawdbot on a Mac Mini, but it can be ran from anything.
It’s like Siri on steroids, and is useful when you want AI to remember things, monitor activity, and take action without constant prompting. It’s less about one-off answers and more about having an assistant that operates in the background.
A big reason Clawdbot has gotten so much attention is because it flips the normal AI flow. Instead of you always going to the AI, Clawdbot can message you first when something matters. Think reminders, alerts, summaries, or updates delivered directly to your phone or chat app.
A few practical, business-friendly examples:
Acting as an inbound assistant that flags urgent emails, messages, or tasks while you’re busy
Sending daily or weekly summaries of activity, progress, or reminders without being asked
Handling simple actions like drafting messages, organizing files, or triggering workflows on your machine
Clawdbot is open-source, meaning its code is public and can updated and improved. It remembers context, reaches out when needed, and works inside the apps you already use. It’s still early, but this is a strong glimpse of where personal and business AI assistants are heading.
Learn more below ⬇️
📸 Startup Spotlight
Krea AI
Krea AI - Real-Time AI Image Generation for Creatives.
The Problem: Artists, designers, and creators often face time-consuming workflows when using generative AI. Many image tools are slow, clunky, and require constant re-prompts, making iteration frustrating and creativity harder to scale.
The Solution: Krea AI offers a real-time, canvas-based AI generation tool. Users can draw, erase, and guide the AI live—getting instant feedback on their creative direction. Krea also supports uploading references and using visual masks to control specific areas, blending manual skill with AI speed.
The Backstory: Founded in 2022 by Spanish engineers Alejandro Matamala and Tony Beltramelli (former co-founder of Uizard), Krea began as a research-driven tool for professional creators. It has since evolved into one of the fastest-growing AI image tools, gaining traction for its speed, control, and responsiveness.
My Thoughts: This is cool and another example of where AI is headed. Lower latency in decision making and movement isn’t something that is necessary revered as a huge update, but it’s an important shift over time as a landmark of AI becoming more effective.
“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
If only NVIDIA’s new weather models came out before the big storm! Till Next Time,
Noah on AI