The Archive

User reports Claude Code cache reduced to 5 minutes, forcing token reprocessing and additional fees for long-context codebases.

User commentary on ChatGPT image generator capability and photorealism quality.

Microsoft is rolling out a new Agent Mode inside Office apps like Word, Excel, and PowerPoint this week. Previously described by Microsoft as "vibe working," the Agent Mode is a more powerful version of the Copilot experience in Office that Microsoft has been trying to sell to businesses. "When we first shipped Copilot, foundation models were not powerful enough to use Copilot to command the applications," admits Sumit Chauhan, corporate vice president of Microsoft's Office Product Group. "This meant Copilot was a passive partner in documents: it could answer questions but missed the mark whe...

Unitree announces wheeled variant of G1 humanoid robot, expanding mobility form factors.

Discussion of popular Claude.md prompt file (78.5k GitHub stars) and user workflows.

OpenAI releases GPT-5.5, advancing capability in coding, research, and data analysis with improved speed and performance.

User reports Claude Opus 4.7 (April 17 launch) shows degraded code editing behavior, prompting return to OpenAI Codex subscription.

Google reports 75% of new internal code AI-generated, up from 25% in 2024, signaling enterprise adoption acceleration.

I spent some time few days back comparing Opus 4.6 and 4.7 using my own usage data - just to see how they actually behave side by side. [https://github.com/getagentseal/codeburn](https://github.com/getagentseal/codeburn) it’s still pretty early for 4.7, but a few things surprised me. In my sessions, 4.7 gets things right on the first try less often than 4.6. One-shot rate sits around 74.5% vs 83.8%, and I’m seeing roughly double the retries per edit (0.46 vs 0.22). It also produces a lot more output per call - about 800 tokens vs 372 on 4.6 - which makes it noticeably more expensive. ...

User reports Qwen 3.6 27B practical performance on data science and Python debugging tasks with 24GB VRAM.

User reports Opus 4.7 uses excessive corporate jargon, acronyms, and invented slang in code generation, reducing comprehensibility.

Brief mention of HY-3 preview with no details or link context provided.

Google Cloud CEO Thomas Kurian discusses enterprise agent platforms and Google's cloud infrastructure advantages in the agentic AI era.

OpenAI Codex configuration guide covering personalization, detail levels, and permissions settings.

Tutorial on setting up Codex workspace, threads, projects, and file management for task completion.

Overview of OpenAI Codex as task automation platform enabling tool integration and document/dashboard generation.

Guide to Codex automation features using schedules and triggers for recurring reports and workflows.

Documentation on Codex plugins and skills for tool integration and workflow automation.

Codex use-case compilation covering task automation, deliverable generation, and cross-tool integration.

Getting-started guide for Codex onboarding covering project setup, threads, and initial task completion.

DeepSeek releases DeepEP V2 and TileKernels optimization libraries on GitHub.

GPT Image 2 outperforms Nano Banana Pro in visual generation quality comparison.

User reports Claude Code removed from subscription benefits email after re-subscribing post-Opus 4.7.

So we all got that extra credit a few weeks back, right? $150 in my case. I turned off extra usage until I needed it, which is today. Tried to call the API - no balance. So the extra usage is only via claude code. Thanks for that, Anthropic. EDIT: So I added $25 in API usage, and the text against that balance says: "Your credit balance will be consumed with **API, Claude Code** and Workbench usage. " Jesus wept, Anthropic. Do you seriously ever get someone to sit and think about what you're doing?

[https://www.businessinsider.com/anthropic-trillion-dollar-valuation-on-secondary-markets-2026](https://www.businessinsider.com/anthropic-trillion-dollar-valuation-on-secondary-markets-2026)

OpenAI codex repo models.json updated removing 'latest frontier model' designation from version 5.4, suggesting imminent new model release.

Google announces first Austrian data center in Kronstorf, creating 100 jobs; infrastructure expansion unrelated to Gemma model development.

Open-source Agent-Quest tool visualizes Claude Code agents in real-time across parallel CLI sessions using fantasy-themed 2D UI.

Performance test: Qwen 3.6 27B with speculative decoding achieves 25.53 tokens/sec with 2x speedup on local hardware.

User reports Qwen3.6-27B performs well on Svelte code generation, suggesting viability of open-weights models vs. paid APIs.