Welcome back to LLM Decode 👋
AI agents are attracting eye-watering valuations just as the infrastructure behind our favorite AI tools is showing its limits. Devin-maker Cognition is reportedly nearing a $47B valuation, while ChatGPT, Claude, Grok, and reportedly Gemini experienced rare overlapping disruptions.
Here’s what matters today.
Devin’s Maker Is Suddenly a $47B AI Giant
Cognition, the company behind AI coding agent Devin, is reportedly preparing to raise around $1B at a $47B valuation. Strong investor demand could push the round even higher, with nearly $10B in investor interest reported.
The numbers are moving incredibly fast. Cognition was valued at roughly $26B in May, meaning the proposed valuation would represent an increase of about 80% in only a few months. Its annualized revenue has also reportedly jumped from $492M in late May to more than $900M.
The bigger story is what investors are betting on. Devin doesn't just suggest the next line of code- it can analyze repositories and work through multi-step development tasks, pushing AI coding from “copilot” toward autonomous engineering agent territory.
Why It Matters
AI coding is becoming one of AI's hottest markets. Investors are betting that agents can capture a meaningful share of software-development work.
Agents are shifting from assistance to execution. The value is increasingly in completing workflows, not simply generating answers.
Revenue matters again. Cognition's reported growth shows that agentic AI isn't only attracting hype, companies are paying for it at scale.
Watch next: Expect more AI products to charge based on tasks completed or compute consumed, rather than traditional per-seat SaaS pricing.
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
This guide distills 10 AI strategies from industry leaders that are transforming marketing.
Learn how HubSpot's engineering team achieved 15-20% productivity gains with AI
Learn how AI-driven emails achieved 94% higher conversion rates
Discover 7 ways to enhance your marketing strategy with AI.
ChatGPT, Claude, Grok and Gemini Hit by Rare Overlapping Downtime
On September 3, several of the world's biggest AI services experienced disruptions at roughly the same time. ChatGPT, Claude and Grok had confirmed outages, while Ars Technica reported Gemini was also affected.
The causes weren't necessarily connected. OpenAI said a routing error made ChatGPT and Codex unavailable for some users. Anthropic reported elevated errors across several Claude models, while SpaceX said Grok's problem came from an outage at its Memphis compute center.
Most services recovered relatively quickly, and there's currently no confirmed evidence of one shared failure behind the outages. But having several frontier AI services stumble at almost the same time offered a preview of what happens when businesses become deeply dependent on AI infrastructure.
Why It Matters
AI is becoming business infrastructure. When AI goes down, coding, research, support and content workflows can stop with it.
Don't rely on one provider. Critical workflows should have a second model or manual fallback.
Reliability becomes a competitive advantage. Model intelligence gets the headlines, but enterprises also care about uptime, latency and predictable APIs.
Watch next: Expect more companies to adopt multi-model AI stacks that automatically switch providers when one fails.
4 AI Tools & Community Workflows to Try
Task-to-Agent Workflow: Turn clear, repeatable work into agent-ready tasks instead of using AI only for quick answers. Define the goal, expected output, constraints, and success criteria upfront so coding agents can handle more of the execution independently.
AI Failover Workflow: Build critical AI processes with a backup route. If your primary model becomes unavailable, move the task to another provider or a manual process so research, coding, support, or content operations don’t stop during an outage. Recent disruptions across leading AI services show why this matters.
Agent ROI Tracking: Measure AI agents by outcomes rather than how often employees use them. Track completed tasks, time saved, human corrections, and cost per successful task to identify where agentic AI is actually creating business value. Cognition’s reported revenue growth suggests companies are increasingly willing to pay for AI that performs real work.
AI Resilience Check: Regularly test what happens when the AI behind an important workflow stops working. Map which processes depend on each provider, identify the highest-risk dependencies, and create recovery steps before an outage happens. The goal is to make AI adoption reliable, not just powerful.
That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.