Random Llama
Random Llama
ProductsSolutionsBlogCase StudiesContact
Get a Quote
Weekly Newsletter

Get AI & productivity insights weekly

Privacy-first tools, workflow tips, and early product access. No spam — unsubscribe anytime.

Random Llama Software

Texas-built weird tools and custom web platforms—fast shipping, no creepy tracking, no enterprise bloat.

Links
  • Home
  • Products
  • Case Studies
  • Blog
  • Solutions
  • Credentials
  • Contact
Services
  • Custom CMS
  • Booking Engines
  • Mobile Apps
  • AI Integration
Connect
  • Privacy Policy
  • Terms of Service
  • Cookie Policy

© 2026 Random Llama Software, LLC. All rights reserved. Privacy Policy

Back to Blog
ai-toolsai-business

Anthropic Hit 30B, Google Went Cheap, AI Agents Spook Security

Robert HattalaApril 20, 2026
p>Lot of AI news moved this week. Most of it is noise. A few things actually matter. Here is what caught my eye.

Anthropic Quietly Ate OpenAI's Lunch

Reports this weekend put Anthropic's annualized revenue north of $30 billion. That is a wild number for a company that was bleeding cash not long ago.

OpenAI is sitting around $25 billion annualized and reportedly eyeing a public listing later this year.

Why it matters. The frontier model race is not a single horse anymore. Enterprise buyers are splitting their bets, and Claude is grabbing a big chunk of that spend. When the incumbent gets passed on revenue, the narrative shifts fast.

My take. I have been on the Claude bandwagon for a while, and this tracks with what I see in the wild. Developers and ops teams keep picking it for serious work. OpenAI still owns the brand, but brand does not pay the bills forever.

Google Is Going After the Cheap Seats

Google rolled out Gemini 3.1 Flash-Lite this week. 25 cents per million input tokens. 2.5x faster response times than the last Flash version.

That is not a flagship play. That is a volume play.

Why it matters. When a frontier lab ships a model this cheap and this fast, it squeezes the middle of the market. Smaller providers who were competing on price just got lapped. Teams running high volume workloads have a very good reason to switch.

My take. Cheap and fast beats smart and slow for 80 percent of real world jobs. Summarization, classification, routing, data extraction. None of that needs a $200 per million token model. Google knows this and is going after the volume while Anthropic and OpenAI fight over the top shelf.

AI Agents Are Scaring the Security Folks

The agent wave is here. Tools that browse, click, type, and buy stuff on your behalf. Cybersecurity folks are not sleeping well about it.

The concern is simple. You give an agent access to your email, your calendar, your browser. Now every phishing email and shady web page is a potential attack on your agent, not just on you. Prompt injection goes from lab curiosity to real threat.

Why it matters. Most companies are not ready. The security stack assumes humans make the click. Agents do not read warning banners. They just follow instructions.

My take. I am bullish on agents long term. But if you are deploying them in your business right now without scoped permissions and a human sign off on anything that writes or spends, you are asking for a bad quarter. Read only agents first. Write actions need guardrails. No exceptions.

Bottom Line

The money is moving. The prices are dropping. The risk surface is expanding. Pick the tool that fits the job, keep a short leash on anything that can act for you, and do not believe your own hype.

Related posts

AI Agents Show Up for Real Work and MCP Gets Cracked Open

Three AI stories worth reading this week. EY put agents on 130K auditors, OpenAI aimed at drug discovery, and Anthropic MCP spec took a beating.

April 19, 2026

Snap Cuts 1,000 Jobs Blaming AI While Amazon Plays Doctor

Snap blames AI for cutting a quarter of its headcount, Amazon rolls out a free Health AI agent for Prime, and lawyers keep filing briefs full of AI hallucinations.

April 18, 2026

AI This Week: Amazon's Doctor, Grok's Honesty, Google's Trick

Amazon launches a health AI agent, Grok 4.20 takes hallucinations seriously, and Google cuts memory bloat. Grown-up AI news from a good week.

April 16, 2026
All posts