OAI Jalapeño fried entire Nvidia stack?? 🌶. .
Free AFIRE Guide | AI Academy | Advertise | AI Mastery A-Z
Plus: Ranking 5 Best AI Video Generator Apps on Your Phone (2026 Updated List)
How to get unlimited Claude limits: 1. Get hired by Anthropic 🤡 Isn’t it easy peasy!? Anyone here actually works there and can confirm the limits are actually unlimited?
What’s on FIRE 🔥
IN PARTNERSHIP WITH SECTION
The State of AI Proficiency in the Enterprise – July 7 | Free Virtual Event
AI is getting smarter, faster, and better – but recent research shows that many enterprise leaders don’t even know what an AI agent is.
Join Section on July 7 from 1-2 pm ET to hear what data says about AI proficiency in the enterprise.
AI INSIGHTS
🌶️ OpenAI Unveils Its First Custom Inference Chip Built by Broadcom
OpenAI unveiled “Jalapeño,” its very first custom-built inference processor designed and manufactured by Broadcom to reduce its reliance on standard Nvidia GPUs:
-
Jalapeño is an AI accelerator engineered specifically for running pre-built models in response to user prompts.
-
OpenAI revealed that its own AI models actually assisted in the development and design of the chip.
-
While the processor is still in the testing phase, early benchmarks show significantly better performance-per-watt compared to current state-of-the-art alternatives.
-
OpenAI can now optimize every single layer of its system, from the chip architecture up to the actual deployment and product experience.
The partnership with Broadcom was first locked in back in October. During a podcast discussion, OpenAI President noted that the company has been targeting specific workloads to build hardware capable of accelerating what’s possible.
Right now, OpenAI is highlighting Jalapeño’s low operating costs, particularly when running real-time coding models. Heavy pre-training workloads will likely still depend on Nvidia hardware for the foreseeable future, but trimming down the day-to-day costs of running inference is a massive lever for the company’s bottom line.
PRESENTED BY HUBSPOT
The ones showing up in LLMs convert 3× better than Google
They optimized for LLMs, not just Google.
FAQs. Comparison pages. Transparent pricing. LinkedIn presence. These aren’t vanity plays. They’re what gets you cited in ChatGPT, Gemini, and Claude when your buyers are researching, your investors are looking, and your future hires are deciding where to work.
Download the free AEO Playbook for Startups from HubSpot and get the exact checklist. Five minutes to read.
AI SOURCES FROM AI FIRE
1. Free: Complete Guide to 5 Levels of Building Your Claude AI Second Brain from Scratch. Turn your scattered information into a fully organized AI brain. Organize notes, transcripts, and projects into a system that remembers everything and finds exactly what you need.
2. Pro: Vague Claude Answers? These 5 Prompts Will Change That Instantly. Learn the 5 prompt techniques that consistently make Claude give clear, precise, and useful answers. It’s your method, these tricks let you control output quality fast.
3. Pro: Claude Cowork + NotebookLM Is Overpowered. It Completely Changed How I Do Research. Here’s an AI workflow automation system that collects research, updates notebooks, and creates slide decks, reports, mind maps, and audio briefings for you.
PRO: 5 BEST AI VIDEO APPS ON PHONE
📱 Ranking 5 Best AI Video Generator Apps on Your Phone (2026 Updated List)
You can NOT choose an AI Video App just because of video quality. You need to see the real experience of using it on your phone.
An app can have a strong model, but if the interface feels messy, slow, or keeps asking you to log in again, you will stop using it.
So I tested 5 popular AI Video App options myself, directly on my phone, without using a laptop. I checked these things:
|
What I checked |
Question I asked |
|---|---|
|
App opening speed |
Does it open fast, or block you with pop-ups first? |
|
Finding features |
Are the tools easy to find, or hard to locate? |
|
Finishing a video |
Can you go from start to finish without frustration? |
|
Output quality |
Is the final video actually good? |
The 5 apps I tested: Runway, Kling, ImagineArt, Hailuo, and Flow.
Some results surprised me. One app looked strong on paper but felt annoying to use on a phone. Another app, less famous, ran smoother than some of the bigger names.
Below, I’ll show you real videos I made with each app, so you can compare them yourself instead of just trusting my words.
TODAY IN AI
AI HIGHLIGHTS
🎨 Figma just made the designer-engineer handoff way easier. You can now pull code into the canvas, test flows, add motion, create plugins with AI prompts (see its demo)
💸 Some of us, as employees, were told to use AI more. Now, companies are trying to cut back after token costs started piling up fast on tiny tasks!? (404 Media reported)
⚠️ The U.S. is pressing Meta to submit its AI models for government review. OpenAI, Anthropic, Google, Microsoft, xAI have already agreed to similar reviews (by Reuters)
📱 Facebook is testing a standalone AI companion app for creators. It gives content ideas, performance insights, and daily priorities inside a rebuilt Creator Studio
(via Facebook blog)
👀 Anthropic accused Alibaba of a massive ‘brazenly’ & ‘illicitly’ AI distillation attack. Basically, using Claude outputs to train smaller models using 25,000 fake accounts.
(CNBC reported)
📊 Plot twist: AI was supposed to kill engineering jobs, but new data suggests they’re the most resilient, making up 55% of new hires. Faster coding just creates more work.
(Challenger’s Job Cuts Report)
🧠 Google just lost 2 more key Gemini researchers to Anthropic. Jonas Adler & Alexander Pritzel are reportedly leaving, adding to a growing AI talent drain.
(Bloomberg reported)
💰 Big AI IPO Watch: Agility Robotics plans a $2.5B SPAC deal to go public, while its Digit humanoid robot already operates across 9 customer sites and has secured $300M+ in multi-year orders.
NEW EMPOWERED AI TOOLS
-
🚀 Cloudways’s Agency Advantage shows you how to build an AI-powered agency that scales without hiring more people. Join 20+ experts in this FREE 2-day event. June 30 and July 1. Claim your spot.
-
🎥 Seedance 2.5 is designed for campaign-ready outputs across vertical, square, and landscape formats suitable for paid ads / social posts.
-
📊 Propane gives your product team and agents one connected, always-current view of your customers. You just build products people love.
-
⚡ Tencent EdgeOne is an edge platform for modern web apps and AI agents. Build with your preferred frameworks and deploy.
-
💳 Stripe.Directory is a single discovery layer for developers and AI agents to find businesses on Stripe and integrate them instantly.
AI BREAKTHROUGH
🖥️ Gemini 3.5 Flash Just Shipped Built-In Computer Use
As of this week, “computer use” is no longer just a standalone, experimental feature, it is now natively integrated as a built-in tool inside the main Gemini 3.5 Flash model:
-
Computer use transitions from the standalone Gemini 2.5 model directly into 3.5 Flash, alongside existing tools like Search and Maps grounding.
-
The model is specifically tuned for long-horizon enterprise tasks.
-
Gemini 3.5 Flash has undergone targeted adversarial training specifically focused on computer use.
-
DeepMind released 2 optional enterprise safeguard systems. The first requires explicit human confirmation before executing sensitive or irreversible actions. The second automatically halts the agent’s workflow if an indirect prompt injection is identified.
While the model itself has built-in protections, DeepMind is explicitly advising engineering teams to combine these capabilities with secure sandboxing, strict access controls, and human-in-the-loop verification.
For teams ready to get their hands on this, the rollout is already live. You can start building with the new computer use capabilities directly via the Gemini API and the Gemini Enterprise Agent Platform.
We read your emails, comments, and poll replies daily
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
Leave a Reply