🌟 Editor's Note: Recapping the AI landscape from 05/05/26 - 05/11/26.
🎇✅ Welcoming Thoughts
Welcome to the 42nd edition of KPAI Weekly.
What’s included: A weekly winner, AI industry impacts, Interview highlight, and Startup Spotlight.
Trying something new this week.
Current plan is for the last issue of each month to be full, but the intermittent issues to be lighter.
xAI and Anthropic teamed up.
We got a humanoid robot update from NEO.
Inaugural ChatGPT futures class of students.
Anthropic has found its compute. Usage is back!
NVIDIA doing NVIDIA things.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: OpenAI // ChatGPT
OpenAI wins an enterprise-heavy week. Between a $4B deployment arm, a new default model, and a self-serve ads platform, OpenAI strengthened its lead over the rest of the field (sans NVIDIA). Here's the recap:
$4B Deployment Company: OpenAI launched a majority-owned deployment arm backed by $4B from TPG, Goldman, SoftBank, McKinsey, and Capgemini. Also acquired Tomoro for ~150 forward-deployed engineers. Good move. Used to help companies implement AI. Just like KPAI just way bigger.
GPT-5.5 Instant Default: New default ChatGPT model with 52.5% fewer hallucinations in high-stakes medicine, law, and finance prompts. Added "memory sources" so users can see what context shaped a response. Cool. I tend to use ‘thinking’ when using GPT.
Self-Serve Ads Manager: CPC bidding beta launched. Ads pilot expanded from US to UK, Japan, South Korea, Brazil, and Mexico. Targeting $2.5B in 2026 ad revenue. Ads in ChatGPT are officially here. Makes sense for GPT. Looks like access is expanding.
OpenAI also shipped voice models in the API - Realtime-2, Realtime-Translate (70+ languages), and Realtime-Whisper. They also added “Trusted Contact,” which notifies a designated person if the system detects self-harm risk.
From Top to Bottom: Open AI, Google Gemini, xAI, Meta AI, Anthropic, NVIDIA.
⬇️ Field Update
Who’s moving, who’s stalling, and who’s climbing: Ordered by production this week.
🟠 Anthropic // Claude
SpaceX Deal + Akamai $1.8B: Anthropic took over SpaceX's Colossus 1 data center in Memphis - 300+ MW and 220K GPUs. Immediately doubled Claude Code and API limits. Separately signed an $1.8B cloud deal with Akamai. Capacity is no longer the bottleneck. Awesome! Interesting they partnered with SpaceX. Usage should no longer be affected by lack of compute.
⚪️ NVIDIA
$5.6B Infrastructure Week: $3.4B deal with IREN for 5 GW of AI infra plus a $2.1B equity stake. $3.2B Corning partnership for US optical capacity with three new plants. Total 2026 equity commitments now past $40B. NVIDIA is the AI economy's banker heading into May 20 earnings. They keep investing, I like it and why not, they have the capital!
🔴 xAI // Grok
SpaceXAI Rebrand: Musk dissolved xAI as a separate company and folded it into "SpaceXAI" under SpaceX. Same day: signed the Anthropic compute deal, launched Connectors, and released Imagine Quality Mode. I’m hoping this move means the AI arm continues to be used more for science and robotics than social media and image gen.
🟣 Google // Gemini
AlphaEvolve Impact Update: DeepMind's Gemini-powered coding agent cut variant-detection errors 30%, pushed a power-grid feasibility rate from 14% to 88%, and improved disaster prediction by 5%. Quiet week otherwise - saving it for I/O on May 19. Sometimes we forget about all the cool projects Google and Deepmind work on outside of Gemini.
🔵 Meta // Meta AI
AI Age Detection via Bone Structure: Meta will use AI to scan photos for height and bone structure to identify under-13 users and remove them. Explicitly "not facial recognition." Lands weeks after a $375M New Mexico verdict on child safety. 8,000-person layoff hits May 20. Interesting tech, smart move, should help them stay out of court.
🤖 Impact Industries 🎓
Robotics // NEO Factory
1X Technologies opened America's first vertically integrated humanoid robot factory — a 58,000 sq ft facility in Hayward, California, purpose-built for their NEO humanoid. First-year capacity is 10,000 units and the entire run sold out in five days at either $20K per unit or $499/month. Consumer shipments are planned for later this year. Finally some news on the NEO robot, the first humanoid we covered!
Read the Story
Healthcare // Drug Discovery
The UK-led OpenBind consortium released its first publicly available dataset and predictive AI model for drug discovery. The model predicts how drug molecules interact with protein targets, and unlike the closed-source competition, the data and weights are fully open. Drug discovery has been one of AI's most promising but most gatekept applications. Free access to production-quality binding prediction could accelerate early-stage development significantly.
Read the Story
💻 Interview Highlight: Jamie Dimon (In Good Company Pod)
Interview Outline: Jamie Dimon discusses the relentless drive required to build a corporate culture that operates like a "neural network" rather than a rigid hierarchy. He details his philosophy on fighting bureaucracy by empowering small, "Navy SEAL" style teams to handle mission-critical technology. Dimon also warns of the systemic risks posed by geopolitical instability and advanced cyber threats like "Mythos," while predicting a future where AI-driven productivity leads to a standard 3.5-day work week.
About the Interviewee: Jamie Dimon is the Chairman and CEO of JPMorgan Chase. He is a veteran leader in the global financial sector, recognized for his "fortress balance sheet" approach and his commitment to operational excellence and transparency.
Interesting Quote: "With more productivity... human beings in the developed world will probably be working three and a half days a week."
Condensed Interview Highlight — Jamie Dimon (In Good Company Podcast)
1. Nicolai: You’ve said that technology teams need to operate like "Navy SEALs." How does that work in practice?
Jamie Dimon: You have to have small teams that are fully authorized and dedicated to accomplishing something. If a project is only 1% of each person's job across departments, it will not get done. You need a dedicated team that has everything they need—the lawyer, compliance, and engineers—directly on that team so they aren't waiting six or nine months for reviews.
2. Nicolai: How do you stop bureaucracy from creeping into such a large organization?
Jamie Dimon: Bureaucracy, complacency, and arrogance will take down a company. I fight it by ensuring all information is shared beforehand; if it isn't, I cancel the meeting. Meetings should never end by saying "we'll pick it up next week"—they must end with a specific person authorized to take action.
3. Nicolai: How is AI currently transforming the way JPMorgan Chase actually works?
Jamie Dimon: We aren't putting our heads in the sand; we’ve been doing this for 13 years and have thousands of people deployed in it. We have use cases across risk, fraud, marketing, note-taking, and prospecting. It should be part of every business review—talking about your projects, your competition, and how you protect yourself from bad guys using the same tools.
4. Nicolai: If AI makes us so much more productive, what does the future of the work week look like?
Jamie Dimon: Sometime down the road, with more productivity and capital formation, I think people in the developed world will probably be working three and a half days a week. We’ll have wonderful lives with a GDP per person around $200,000. However, unlike the internet or steam engine, this shift may happen much faster than the ability of our society to adjust.
5. Nicolai: How do you manage the relationship with your board to ensure they are actually coaching you?
Jamie Dimon: My board meets 100% without me at every single meeting. I have senior executives come in and tell the board what they think—they are encouraged to disagree with me on risk or people decisions. Afterward, the lead director gives me a call or a note with coaching on what I could do differently and what they’re concerned about.
📧 Startup Spotlight
Slashy AI
Slashy — AI for email that sounds like you.
The Problem: You spend half your day trapped in your inbox, but you can’t just outsource it to a generic chatbot because it sounds like a robot. Traditional AI "slop" is easy to spot—it’s too formal, too wordy, and doesn't know your specific context. To keep your relationships real, you’re forced to manually type every single "Sounds good, let's circle back Tuesday."
The Solution: Slashy is an AI-native email client that actually learns your writing style. It doesn't just "draft a reply"; it studies your past sent folder to understand your tone, your specific sign-offs, and how you actually talk. It triages your inbox, pulls context from your CRM and notes, and prepares replies that look like you wrote them. You aren't "using AI"—you're essentially giving your email a brain that thinks (and writes) exactly like you do.
The Backstory: Built by the YC S25 team (including 18-year-old prodigy Pranjali Awasthi and Harsha Gaddipati), Slashy was born out of frustration with "AI-bolted-on" tools. They realized that for an AI assistant to be useful, it needs a perfect memory and a perfect mimic. They’ve raised $81M to make the "human-in-the-loop" experience feel less like editing a robot and more like approving your own thoughts.
My Thoughts: I like this and have actually had clients ask for something similar in the past. Cost seems reasonable and it comes with full API / MCP access to your system. Will likely give it a try and add it to the KPAI Audit toolbox.
“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
Till Next Time,
Noah from KPAI