- AI Fire
- Posts
- 🏦 Nvidia Goes Debt Shopping
In partnership with
Google DeepMind just mapped out exactly how we get from AGI to ASI, it turns out the future looks more like millions of digital workers cooperating together.
What's on FIRE 🔥
IN PARTNERSHIP WITH SECTION
If you still don’t know how to demonstrate AI ROI at work, come steal this reporting framework.
On 6/17, from 12-1pm ET, join this free webinar to learn the 3 ways every leader should be thinking about AI value.
You’ll leave understanding core success metrics at the company, team, and project level.
AI INSIGHTS
Table 3 | Major technological pathways from AGI to ASI
That’s exactly what Google DeepMind mapped out in a new paper detailing the leap from human-level AGI to Artificial General Superintelligence (ASI). It turns out that building superintelligence is about navigating very real, physical roadblocks.
DeepMind sees ASI emerging through a mix of 4 routes: brute-force scaling, algorithmic paradigm shifts, recursive self-improvement (AI writing better AI), and massive multi-agent collectives.
The most viable path to ASI isn’t a single monolithic model. Instead, picture millions of AGI-level digital workers cooperating at high bandwidth.
Current models are incredible at remixing human language, but they can't easily discover entirely new physics from raw data. True ASI needs to generate novel conceptual primitives without human hand-holding.
Even if an ASI thinks a million times faster than us, it still has to run physical experiments to validate new science. You simply cannot speedrun a chemical reaction or real-world material testing.
To keep capabilities climbing, DeepMind points out that labs will have to shift massive compute toward test-time reasoning and high-fidelity physical simulators. The transition to ASI won't be a sudden, magical overnight explosion.
PRESENTED BY HUBSPOT
HubSpot AEO
Picture this. A buyer opens ChatGPT and asks for a recommendation in your category. Your competitor's name comes up. Yours doesn't. And that buyer never makes it to your website.
That's happening right now in markets everywhere. And most teams don't know it's happening because it never shows up in their analytics.
HubSpot AEO shows you exactly where your brand stands in AI search, where competitors are getting recommended instead of you, and tells you specifically what to fix. No expertise needed.
Try it free for 28 days. Just $50 a month after.
AI SOURCES FROM AI FIRE
1. Claude + ChatGPT + AI Agents: A Step-by-Step System to Actually Make Money With AI. You’re using AI every day but the money is not coming? Here’s the exact system that turns prompts and workflows into real, measurable revenue.
2. Free Guide: How to Use Gemini Omni: 5 Mind-Blowing & Practical Use Cases Most People Miss. Gemini Omni edits real footage, adds drone movement, translates speech, and places 3D text in your scene. 5 use cases most creators have never tried.
3. Video: HTML Is the New Markdown for Claude. Markdown worked… until AI got too powerful, and feeling limiting. Now I’m leaning on HTML, and so is the Claude Code team, for richer visuals, interactive outputs, and easier sharing.
BEST AI AGENT TOOLS IN 2026
Have you ever wondered which AI agent tools actually deliver?
I’ve tested every major tool from A to Z, putting them through real-world tasks to see which ones can actually handle the job and which are just flashy disappointments.
We put 3 of the most talked-about tools to a direct test: n8n, Claude Code, and Base44:
👉 Build an AI-powered task manager where anyone can add tasks, mark them done, and get an AI-generated priority ranking.
By the end, you’ll know exactly which tool fits your skill level, what each one actually produces, and which one gives beginners the clearest path from idea to working product.
TODAY IN AI
AI HIGHLIGHTS
🚦 Before adding another AI tool & model to your workflow, read this new guidance on when not to use AI first. If you can’t validate it, audit it, or explain it, don’t trust it.
✅ Want unreleased Gemini features before anyone else? Google is opening a limited number of slots for users to join the Gemini Trusted Tester program. Sign up here.
📋 Job interviews are becoming AI tests. Your new interview flex might be how well you use Claude, ChatGPT, or another AI tool to think through a real work problem.
🛡️ A group of 76 cybersecurity veterans published an open letter to the U.S., urging the government to reverse its ban on Anthropic’s Fable 5 and Mythos models.
🔎 Meta is rolling out AI Mode on Facebook. It answers search questions using public posts from Groups, Reels, and more. Useful… until it summarizes bad group advice.
🧪 A 13-word Reddit comment might be enough to trick AI search. New Cornell research says these tiny snippets can quietly shape what ChatGPT & Google AI say.
♻️ Google researchers are turning 2,000 retired Pixel phones into a mini datacenter. The goal: give old phones a second life & cut the carbon cost of buying new servers.
📈 Nvidia plans to raise at least $20B through investment-grade bonds since the AI boom began. It’s doing this even while generating $49B in free cash flow last quarter.
💰 Big AI Acquisition: Salesforce is buying Fin for $3.6B to strengthen Agentforce. Fin’s AI agent handles support across chat, WhatsApp, SMS, calls, Slack, and more. The deal should close in early 2027.
NEW EMPOWERED AI TOOLS
📈 Agorapulse's Scorecard helps marketing teams predict & validate winning content before creating it, kill weak ideas instantly with a FREE 4-step content validation system. Get Free template.
🔗 Novu Connect ships agents where your users already work across Slack, Teams, WhatsApp, Telegram,… without building each integration.
💼 Wobo is your personal AI recruiter. It matches pre-vetted jobs to your background, then applies on each company's own site for you.
💸 Kickbacks.ai helps you get paid for Claude Code wait states. Advertisers bid for a tiny sponsored status line; users get 50% of ad revenue.
AI BREAKTHROUGH
OpenRouter dropped a new API called Fusion, and it is proving that teamwork actually makes the dream work in AI. By pooling responses from multiple models together, they are nearly matching the heavyweights at a fraction of the cost:
How it works: Fusion sends your prompt to several models simultaneously. A separate evaluator model reviews all the generated answers and merges the best parts into one final, highly accurate reply.
A tag-team of DeepSeek V4 Pro, Kimi K2.6, and Gemini 3 Flash scored 64.7% on a Perplexity benchmark. That is just shy of Fable 5’s 65.3%, but it costs about half as much to run.
OpenRouter CEO is betting big against single-model dominance, stating, “The future of AI is neurodiversity, not single-model takeovers.”
The concept of a "model council" isn't entirely new, Perplexity & Grok already use similar routing systems to optimize their outputs. But OpenRouter takes that complex architecture and wrapping it up into an easy-to-use API for everyday devs.
With the recent restrictions surrounding Fable, builders are being forced to get creative to maintain access to frontier-level performance. Fusion gives developers a viable, cost-effective alternative to relying on a single mega-model.
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