- AI Toast
- Posts
- Claude Opus 4.6 is here and it’s thinking
Claude Opus 4.6 is here and it’s thinking
Plus: Augment Code launched Context Engine as an MCP server
Poonam Soni
February 06, 2026
Welcome to AI Toast!
Here’s what’s on the menu today:
Claude Opus 4.6 is here and it’s thinking
Super Bowl snark sparks AI Ad war
Reddit revenue spikes
Augment Code launched Context Engine as an MCP server
Superhero Tools to build websites & apps
Tutorial on how to create slides with Kimi
Quick Toasts: Sam’s response to Anthropic’s ads
Total read time: About 5 minutes, perfect for a quick coffee break.
Anthropic just released Claude Opus 4.6, and it’s effectively a senior engineer that doesn’t sleep or forget your instructions. This update focuses entirely on sustained reasoning and managing complex, multi-day projects without losing the plot.
Adaptive Thinking: This new mode lets the model decide how much "thought" a problem requires before answering, replacing rigid token budgets with dynamic reasoning.
128k Output Tokens: You can now generate significantly longer code files or documents in a single go, doubling the previous limit of 64k.
Compaction API: A new beta feature automatically summarizes older parts of a conversation to keep them in context, effectively allowing for infinite project threads.
Agentic Coding: It tops the Terminal-Bench 2.0 scores, capable of planning, coding, and debugging its own work better than any previous iteration.
Enterprise Availability: The model launched immediately across AWS Bedrock, Google Vertex AI, and Microsoft Foundry, signaling a heavy push into corporate workflows.
Context Window: It now supports a 1 million token context window in preview, allowing it to hold entire repositories or legal archives in active memory.
Opus 4.6 is built to handle the "boring" middle parts of long projects that usually trip up AI agents.
Personal Take: The real-life benefit here is the end of "amnesiac" AI interactions. With the Compaction API and massive context, you can finally build a digital partner that remembers the why behind a project weeks after you started it. This shifts your role from constantly re-explaining context to simply guiding a capable teammate who knows the history as well as you do.
Image credits: Reuters
Reddit just proved that AI isn't just for chatbots and it is for making money. The company smashed earnings expectations largely because its new AI ad tools are actually working.
Revenue Jump: Q4 revenue hit $726 million, a 70% increase that comfortably beat analyst estimates of $665 million.
AI Ad Engine: The boost came from new AI features like "Max" campaigns, which automate ad bidding, and tools that generate copy or crop images for advertisers.
Strong Forecast: They aren't slowing down, predicting Q1 2026 revenue will land between $595 million and $605 million, marking another 50% year-over-year jump.
User Growth: Daily active users climbed to 121 million, proving that people are still showing up for human conversation despite the AI noise.
Buyback Program: Confidence is high enough that the board authorized a $1 billion share repurchase program to return capital to shareholders.
Search Play: "Reddit Answers," their AI search feature, grew to 15 million weekly users, signaling a shift toward becoming an answer engine, not just a forum.
This performance shows that Reddit has figured out how to monetize its data without killing the user experience. The market responded well, with shares ticking up on the news.
Personal Take: The most interesting piece here is "Reddit Answers." If Reddit can successfully pivot from a messy forum to a clean, AI-curated answer engine, they solve their biggest historical problem of accessibility. I expect them to start putting premium ads directly inside those AI answers by the end of the year, effectively creating high-value real estate that bypasses the traditional feed entirely.
In partnership with Augment Code
Augment Code's Context Engine employs semantic indexing and mapping to comprehend flow, going beyond simple grep or keyword matching.
With the new MCP, teams can integrate Augment's Context Engine with their preferred agents. This integration should deliver faster and more comprehensive results.
Using the Context Engine is expected to reduce token usage and potentially lower costs.
It is available through remote MCP or local MCP with the Auggie CLI.
Supported agents include: Claude Code / Cursor GitHub Copilot / Gemini CLI
Claude Code with and without the Augment Context Engine's MCP server.
On the left you see Claude Code with Augment Context Engine which finished in 30 seconds vs on the right Claude Code without the engine which took 44 seconds.
Go to Context Engine MCP
Install the MCP server for your agent and after one time indexing, your responses will be quicker and more accurate.
Superhero Tools (Build websites & apps)
Google Opal: This no-code builder turns your text instructions into functional mini-apps and workflows, effectively letting you automate daily tasks without waiting for a developer.
Lovable: This platform lets you build complete web apps and websites just by chatting, handling both the design and the backend code in the background.
Google AI Studio: This web-based prototyping environment gives you free, direct access to Google’s Gemini models so you can test prompts and build features without setting up a complex local environment.
Emergent: This tool uses coordinated AI agents to plan, code, and deploy full-stack web and mobile applications from a simple description.
Bolt: This browser-based sandbox allows you to prompt, run, edit, and deploy full-stack web applications instantly, effectively replacing your local development environment with a single URL.
OpenAI CEO Sam Altman delivered an unusually public and forceful rebuke of rival Anthropic’s Super Bowl advertising campaign, laying bare deep divisions over advertising ethics, monetization and competitive strategy in artificial intelligence.
The February 2026 dispute erupted after Anthropic aired four pointed commercials that openly satirized OpenAI’s plans to introduce advertising into ChatGPT’s free tier, prompting Altman to respond with a lengthy, emotionally charged critique accusing Anthropic of bad faith and authoritarian thinking.
Anthropic lit the fuse with Super Bowl ads mocking ChatGPT’s planned in-chat advertising; portraying AI that drops awkward promos into deeply personal conversations.
OpenAI CEO Sam Altman didn’t laugh for long. He fired back with a blistering public thread, accusing Anthropic of fear-mongering, elitism and “authoritarian” thinking.
At stake is more than marketing shade, it’s a fight over how AI should make money without breaking trust.
OpenAI argues ads are the price of global access, Anthropic insists Claude will stay ad-free, ethics over scale.
The clash exposes a core rift in the industry as AI races from research to revenue, with regulators watching closely and users caught in the middle.
Anthropic’s ads are intentionally alarmist, but they’re not imaginary. They’re a caricature of a real tension that OpenAI hasn’t resolved yet.
Personal Take: Altman’s reaction suggests the ads landed closer to the truth than he’d like to admit. When CEOs stop brushing off satire and start writing manifesto-length responses, it’s usually because the joke hit something real. This isn’t really about who’s being more ethical. It’s about who gets to define the future of AI monetization and who users should trust once money starts talking inside the conversation itself.
Quick AI News Bites
The AI wars escalated quickly this week as OpenAI launched a new agentic coding model just minutes after Anthropic released theirs. It was a clear signal that neither company plans to let the other have a quiet press day. This rapid-fire timing shows just how aggressive the race for developer loyalty has become.
Meta is apparently testing a standalone space as it considers giving its AI feed an app of its own to house viral generated content. This move would separate the endless stream of AI images from your main social feeds. It suggests they know users might want a break from the "slop" mixed in with their friends' photos.
The tension between the labs got personal when Sam Altman reacted to Anthropic's ads by calling them "funny but deceptive." Anthropic’s marketing campaign explicitly targeted ChatGPT users by highlighting where their model falls short. It is a rare moment of public friction between the usually polite industry leaders.
Google proved its scale advantage as the Gemini app crossed 750 million monthly active users globally. This massive number indicates that their aggressive integration into Android and Workspace is paying off. It puts them in a dominant position purely based on how many people interact with their AI every day.
Executives remained tight-lipped during the earnings call as Alphabet refused to discuss the Google-Apple AI deal despite direct questions from investors. The silence suggests the terms of their partnership are either still in flux or strictly confidential. Wall Street is eager to know how heavily the next iPhone will rely on Google’s infrastructure.
Boost revenue and gain new customers by partnering with us
Reach over 35K AI enthusiasts with your product.
Join our newsletter to connect with tech professionals, investors, engineers, managers, and business owners worldwide. DM now!
Got feedback, a story worth toasting, or a wild tech question? I’d love to hear from you; just reply, or find me on X.
Cheers,
— Poonam Soni