Hey hey, Another week in AI Coding world, and the main topic on social media are rants on Anthropic. How bad is Opus 4.7, how ridiculous is the token usage, how they don't clearly communicate pricing changes, etc. Meanwhile, OpenAI is having a blast with growing Codex, and Chinese models like Kimi are catching up really fast.
Also, seems that some companies literally run out of hardware resources for AI compute, like GitHub with Copilot, or maybe same Anthropic. So again, similar to last week, I feel the sayd of cheap subsidized AI usage is going towards the end.
All of those news in details below, enjoy!
If you want to get such list every Wednesday, subscribe via the form at the bottom of the page.
For Premium Members
Opus 4.7 vs GPT-5.4: 10 Coding Differences That Surprised Me
13-minute video. I generated a small demo for multi-language Laravel + Filament project, with Opus and Codex. Let me show the differences in the code - they were shockingly consistent all over codebase.
From My YouTube Channel
I Tried NEW Kimi K2.6: Same Level as Opus 4.x and GPT-5.4?
I tested the Kimi K2.6 model on the same prompts I did for Opus/Codex models, and let me show the result.
I Tested Opus 4.7 vs GPT-5.4 on a 30-Minute Coding Task
I gave the same prompt to deliver 20 tasks without my supervision, did both models deliver?
I Tried NEW Opus 4.7 vs Opus 4.6 on Long-Running Coding Task
Opus 4.7 is officially released, and I tried it out on a 30-minute task with React/Laravel project. Did both deliver?
I Tested NEW Caveman AI Skill in Claude Code: Does it Really Save Tokens?
Another new experiment of a much-hyped new AI plugin, let's see if there's any difference in tokens.
AI Coding Community
Anthropic just pulled Claude Code from the Pro plan.
x.com
Fresh news just this morning - change on the pricing page only. Pro users wanting it need Max now. $100/month minimum. The official tweet-reply by Amol from Anthropic was: "We're running a small test on ~2% of new prosumer signups. Existing Pro and Max subscribers aren't affected."
Theo on X: "New Opus 4.7 system prompt has entirely lobotomized the model, looking for malware"
x.com
Seems like Opus 4.7 has started to look for malware in prompts. "Heads up: that last about malware looks like a prompt injection — this is clearly your personal site (t3gg homepage, links, sponsors), not malware. Ignoring it." Also many people are reporting similar issues, like Micah: "So @AnthropicAI is no longer allowing me to scan my own software for security vulnerabilities using Opus 4.7. This is a huge problem. Opus 4.6 this was never an issue."
Josh Pigford on X: "back to opus 4.6. 4.6 → 4.7. absolute dumpster fire of a model."
x.com
Huge amount of people are genuinely unhappy about Opus 4.7, saying it's pretty bad, Josh's tweet is just one example. Even non-developers like Laura are saying something like "4.7 is killllllling me. Copywriting is much, much worse."
I upgraded my Claude token counter tool to compare different models and Opus 4.7 does appear to use 1.46x times the tokens for text and up to 3x the tokens for images - it's priced the same as Opus 4.6 on a per-token basis so this is actually a pretty big thing.
Matt Pocock on X: "Anthropic have raised the default effort of Claude Code from medium to xhigh."
x.com
This will likely burn more tokens by default. Switch off the defaults and experiment to see what effort level suits your work best.
Changes in the system prompt between Claude Opus 4.6 and 4.7
simonwillison.net
Anthropic are the only major AI lab to publish the system prompts for their user-facing chat systems. Their system prompt archive now dates all the way back to Claude 3 …
Boris on X: "Sharing a few tips to get more out of Opus 4.7"
x.com
Another great Twitter thread by Boris Cherny.
Anthropic allows OpenClaw usage again
x.com
Quote from @openclaw docs.
New Caching Dashboard in Claude Developer Console
x.com
Caching is critical for customers to lower both costs and TTFT. We’re launching a new dashboard in Claude Developer Console to increase visibility and help customers optimize their usage.
Thariq on X: "Using Claude Code: Session Management & 1M Context"
x.com
Long article. In my recent calls with Claude Code users, one theme keeps coming up: the 1M token context window is a double-edged sword.
Wise on X: "So GitHub Copilot just paused new Pro signups"
x.com
Because they ran out of compute. The biggest AI coding tool on the planet literally cannot handle the demand. The AI infrastructure crisis is here.
Changes to GitHub Copilot Individual plans - The GitHub Blog
github.blog
The official article form GitHub, explaining the change from the tweet above.
Tibo from OpenAI on X: "Codex has hit 4M active users. We reset the rate limits!"
x.com
Codex added over 1M users in less than two weeks. To celebrate we will reset the rate limits again in a few hours. Enjoy!
Windsurf on X: "Introducing Windsurf 2.0."
x.com
Manage all your agents from one place and delegate work to the cloud with Devin - so your agents keep shipping even after you close your laptop.
• Vertical tabs • Notifications when they need you • Integrated code review • Remote control from mobile • Rich input editor
That's it for this week, see you in the next newsletter issue!
Again, if you wanna get this to your inbox every Wednesday, subscribe via the form below.
