- AI Fire
- Posts
- 💡 ChatGPT Futures for AI-Native Students
💡 ChatGPT Futures for AI-Native Students
Your AC unit is now an Nvidia data center!
Wendy
May 06, 2026
In partnership with
Your AC unit is about to become a revenue-generating asset. See Span & Nvidia’s XFRA project, which turns residential smart panels into distributed data center nodes.
What's on FIRE 🔥
IN PARTNERSHIP WITH SECTION
On 6/4, join Scott Galloway and Section to learn everything you need to scale from individual productivity gains to enterprise-wise AI orchestration.
Learn how to scale siloed AI use into organization-wide impact and get the playbook to turn your org into a connected, augmented, AI powerhouse.
AI INSIGHTS
Startup Span just announced a partnership with Nvidia & homebuilder PulteGroup to launch XFRA, a distributed data center that lives right next to your AC unit. They’re using their famous "Smart Panels" to orchestrate the whole thing.
Each XFRA node is a total beast with 16 Nvidia RTX PRO 6000 Blackwell GPUs (the liquid-cooled server edition), AMD EPYC CPUs, and 3TB of RAM.
Most homes only use about 40% of their electrical capacity at any given time. Span’s smart panel identifies that spare electricity and funnels it into the XFRA node to run AI tasks for the cloud when your dryer or oven isn't running.
In many markets, hosting a node could mean free electricity and free internet. Span essentially pays your utility bills in exchange for using your wall space and spare power.
They are already rolling this out in new-construction communities with PulteGroup. Instead of one giant 100MW data center, they can deploy 8,000 of these "mini-nodes" 6x faster and at 1/5th the cost.
We’ve been worried about massive data centers sucking the grid dry, but Span is suggesting we just use the capacity we've already built but aren't using. It turns your home from a passive cost-center into a revenue-generating asset.
PRESENTED BY HUBSPOT
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
FREE AI SOURCES FROM AI FIRE
1. AI Agents + Claude Skills Fully Explained: How to Build Your First Agentic Workflow. See how AI agents work beyond prompts. Learn Claude Skills, context management, and build effective agentic workflows with proven techniques.
2. Video: Claude Cowork Tutorial: From Chatbot to Content Assistant Using 3 Simple Workspace Files. Help Claude plan videos, draft hooks, and continue your workflow without repeating context every time.
3. Video: Build Anything with Claude Code in 10 minutes (Zero Coding Required!). How to master Claude Code, the breakthrough "Agentic AI" that doesn't just give advice, but actually executes tasks directly on your computer
CLAUDE DESIGN VS. GOOGLE AI STUDIO
If you’re looking for a AI designer, 2 most powerful models now are Claude Design & Google AI Studio. Both are strong, but they work in completely different ways. So I ran 5 rounds of prompts, from simple to complex, to measure specific areas:
Round 1: Simple landing page → the same fintech brief sent to both tools, no style guidance given, pure default taste on display
Round 2: Pixel-perfect constraints → a 600-word design system brief with exact color tokens, typography scale, cubic-bezier transitions
Round 3: Reverse engineering design language → 3 reference screenshots to extract the design DNA & rebuild it for a different product
Round 4: The stakeholder chaos test → 9 simultaneous complex changes from 3 different stakeholders, sent in a single prompt
Round 5: Emotional brand intelligence → a brief with no colors, no fonts, no layout, only brand character and emotional targets
There’s also a full scoring table across 13 criteria at the end for you, the exact prompts used in every round, copy-paste ready, so you can run the same tests yourself or adapt them for your own projects immediately.
TODAY IN AI
AI HIGHLIGHTS
🎓 OpenAI introduced “ChatGPT Futures Class of 2026”, giving 26 student builders $10K grants & frontier AI access. It's the first gen to start & finish college with GPT.
📊 ChatGPT now works directly inside Excel & Sheets. You can build formulas, analyze data, generate insights or format spreadsheets. Free beta is rolling out for paid users.
📈 Anthropic just launched 10 ready-to-use AI agents built specifically for finance & insurance (here's the list). They can review earnings, build decks, screen KYC files,...
🕵️ Researchers exposed a new AI scam trick: hidden instructions written in Morse code, like dots & dashes. Some agents actually followed them. Here’s how it works.
🍎 Apple agreed to pay $250M to settle a lawsuit tied to claims that iPhone buyers were misled about Siri’s new AI features. Even though Apple admitted nothing.
📱 OpenAI is reportedly building an AI phone that completes tasks, understands goals, works like an operator, not just a normal device. Production may start by 2027.
🛰️ Anthropic partnered with SpaceXAI to secure access to Colossus 1, a massive AI supercomputer. Even crazier: both orgs are exploring orbital AI compute in space.
💰 Big AI Capital Moves: Anthropic committed to spend $200B on Google Cloud & AI chips over 5 years. The deal represents over 40% of Google Cloud’s revenue backlog as the AI infrastructure war explodes.
Traditional SEO is no longer enough. AI evaluates authority, entities, and originality before deciding what to show. If your content doesn’t meet that bar, it gets overlooked.
Join Fernando Angulo (Semrush) and Felipe Bazon (CSO at HEDGEHOG) for a live session on how to adapt your strategy using topical authority, entity SEO, and information gain. Leaving knowing how to:
Master topical authority and control what AI surfaces
Use entity SEO to signal context, relationships, and trust
Create information gain that makes your pages impossible to overlook
Build content ecosystems that perform across search and AI results
Apply winning strategies from real case studies
NEW EMPOWERED AI TOOLS
🕶️ Shadow understands your conversation, tracks what needs to happen, and executes tasks in real time. Our goal is to have no post-call work.
🤖 Superset runs 100s of coding agents on any machine from anywhere. Works for any CLI agent.
✈️ Gyro Autopilot scans your inbox for unclaimed flight money from delays, cancellations, overbookings, and more. No credit card.
⚡ WOZCODE helps spend fewer tokens, finish tasks faster & improve agent performance without switching subscriptions, or workflows.
AI BREAKTHROUGH
Subquadratic just launched SubQ, a model that can swallow 12M tokens in a single pass, that’s like feeding an entire code repository into one prompt. It's faster, it's significantly cheaper ($8 vs $2,600!), and it might just make those complex RAG pipelines a thing of the past.
Their SSA only looks at the token relationships that matter. It runs 52x faster than the industry-standard FlashAttention.
In long-context benchmarks, SubQ ran for just $8, compared to roughly $2,600 on traditional frontier models.
While most models lose the plot as files get bigger, SubQ hit 92% recall at 12M tokens. For context, Gemini 3.1 Pro and GPT-5.4 struggled significantly on similar multi-needle tests (MRCR v2).
They’ve already launched SubQ Code, a CLI agent that can load your entire codebase in one pass. The team is also targeting a 100M-token window by the end of 2026.
We’ve seen "Transformer killers" before (like Mamba), but SubQ feels different because it’s launching with a production-ready API & a world-class team from Meta, Google, and Oxford. If you’re a developer, this is a massive win.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply