- AI Fire
- Posts
- 😰 OpenAI Being Attacked by Meta & xAI
Read time: 5 minutes
AI agents were supposed to change how we work. But the real numbers is way uglier than the hype. Most can’t even finish basic office tasks… Even Gemini 2.5 pro
What are on FIRE 🔥
IN PARTNERSHIP WITH RUBRIK
Rubrik intends to acquire Predibase, a cutting-edge provider of tools for AI model training and deployment and the creator of the open source LoRA eXchange project.
Join July 15 for a live conversation with Predibase Co-Founder and CEO, Devvret Rishi, and Rubrik Chief Product Officer, Anneka Gupta.
AI INSIGHTS
You’ve probably heard the hype: AI agents are the future of work. They'll automate your calendar, write code, send emails, and do your job for you while you nap in a hammock. But they’re barely functioning interns with delusions of grandeur. So… How bad are they?
Researchers at CMU and Salesforce gave leading AI agents some realistic, office-style tasks: write code, search the web, message teammates, follow instructions:
Gemini 2.5 Pro: best performer… completed just 30% of tasks.
Claude 3.5 Sonnet: about 24% success.
GPT-4o: a dismal 8.6% (yes, really).
Other models? Mostly under 10%. Some near 1%.
And we’re talking basic multi-step workflows. The kind of stuff that junior employees do on day two of onboarding.
To make this official, CMU launched TheAgentCompany, a benchmark that simulates a small software company, so you can throw AI agents into a fake startup and see if they survive.
The "Agent" You Bought? Might Not Be an Agent. Gartner calls it “Agent Washing.” Most “AI agents” out there are just fancy AI assistants that can’t plan more than two steps ahead
Gartner estimates only ~130 vendors are building anything close to true agentic AI. The rest is just marketing! They predict by 2027, 40%+ of agentic AI projects will be cancelled.
But Wait, There’s a Bright Spot (Kind Of). Not all agent use cases are trash. Some are decent at code generation (even if incomplete) + workflow automation (but only in narrow setups)…
If you keep them in a sandbox, monitor their output, and keep tasks simple and linear, they can help.
Why It Matters: Everyone’s lying to themselves. Vendors are racing to show progress (even if it’s fake progress). VCs want to fund the “next big platform.” Enterprises don’t want to fall behind, so they over-buy and under-assess. And let’s be honest… the dream of JARVIS is just too good to let go of.
PRESENTED BY BELAY
Economic pressure is rising, and doing more with less has become the new reality. But surviving a downturn isn’t about stretching yourself thinner; it’s about protecting what matters most.
BELAY matches leaders with fractional, cost-effective support — exceptional Executive Assistants, Accounting Professionals, and Marketing Assistants — tailored to your unique needs. When you're buried in low-level tasks, you lose the focus, energy, and strategy it takes to lead through challenging times.
BELAY helps you stay ready for whatever comes next.
TODAY IN AI
AI HIGHLIGHTS
🥊 A X user pitted the top coding models against each other in a fight to the death, with each tried to shut down the others’ processes while staying alive. Here’s the champion.
📸 Higgsfield’s new “Soul” went viral for its “fashion-grade realism” & trendy style presets like Y2K. It’s blowing up across creative circles this week. See some examples.
⚠ 26 YouTube channels pumped out fake, AI videos about the Diddy trial with nearly 70M views across ~900 videos. They turned serious news into clickbait content. Here’s one channel.
🥇 Google launched the full version of Gemma 3n. Explore some of the innovations behind this model, and see how to start building today. Here’s a video overview.
💼 Meta’s reportedly hiring 4 OpenAI researchers for its new super AI team. OpenAI called this a “side quest.” It’s now “recalibrating comp” with the memo to retain staff.
→ Seems like both Elon Musk (xAI), and Mark Zuckerberg (Meta) want to control AI by crushing ChatGPT’s father.
💰 AI Daily Fundraising: Meta aims to raise $29 billion, including $3 billion in equity and $26 billion in debt, from investors like Apollo and KKR to expand AI data centers and deploy 1.3 million GPUs by 2025
AI SOURCES FROM AI FIRE
🔥 Ep 13 tooldrop arrived. And no, it’s not cursed. Unless you hate saving time.
Create, publish high-quality AI-generated books quickly
Generate search intent optimized FAQs for any URL for FREE
Turn your raw & messy data into dashboards and real answers
Build website, write content, capture leads all in 2 minutes
Automate thousands of hours of work in seconds
Note: These exclusive resources & reviews are available only in our AI Fire community. It’s because you guys can freely ask for support or share personal experience during testing there. Get your full breakdown here (no hidden fee)!
NEW EMPOWERED AI TOOLS
⚙️ QuickAgent easily builds AI agents that connect to any service, no-code.
📢 Mintly 1.0 clones polished ads from the biggest companies in the world.
🧠 MyLens turns your raw ideas & content into clear, interactive visuals fast.
🎤 Neura turns your voice notes into actionable content with 20+ AI tools.
🔍 Content Gap sees what content your competitors rank for that you don't.
AI QUICK HITS
🎥 This interview with Microsoft CEO points to exciting paths for AI future.
🤚 If you're using ChatGPT for any of these 11 things, stop immediately.
💵 OpenAI charges by the minute when transcribing, speed up your audio.
🤐 DeepSeek has reportedly pushed back the release of its R2 model, again!
🤖 People now seek 'AI whisperers' to guide through AI's complex world.
AI CHART
Forget what you’ve heard about people falling in love with their chatbots. Anthropic just dropped a fresh report on how folks are actually using Claude, and emotional support? It’s barely a blip.
Anthropic ran a huge study analyzing 4.5 million conversations with Claude using its in-house tool, Clio to figure out how many people are really turning to AI for companionship → Not many.
Only 2.9% of chats even touched on emotional support. And inside that tiny sliver, companionship and roleplay made up less than 0.5%.
Anthropic describes it as a “utilitarian” relationship. Basically, Claude is more like a polite co-worker than a pretend soulmate.
But here’s where it gets interesting...
Even when users do open up emotionally, the conversations usually end on a positive note. Sentiment tends to get better, not worse, as chats go on. No “HER” movie moments.
Still, Anthropic makes it clear: they don’t know if that positivity translates into real-world well-being. There’s no long-term tracking here, just chat vibes.
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