Meta started rolling out Contemplating mode for Muse Spark, where 16 agents will work on your prompt to synthesize a consolidated answer! Deep Think for free? h/t @borrowed_ideas
PROMPT ENGINEERING
-
AI Agent Hits Token Limit Fetching and Parsing PDFs
By
–
and another one! "I'm out of tokens to actually fetch and parse those PDFs in this turn"
-
Claude replaces complex automation workflows with single command
By
–
me at 2AM realizing I wasted 4,000 hours building Zapier-to-Notion pipelines
— Charly Wargnier (@DataChaz) 11 avril 2026
when a single terminal command to Claude literally organizes my entire life https://t.co/ph2uieT9AA pic.twitter.com/LYdqS5hGMWme at 2AM realizing I wasted 4,000 hours building Zapier-to-Notion pipelines when a single terminal command to Claude literally organizes my entire life
-
OpenAI’s Scratchpad for Parallel Codex Chats
By
–
OpenAI is working on a new experimental feature for Codex called Scratchpad.
— 🚨 AI News | TestingCatalog (@testingcatalog) 11 avril 2026
Users will be able to start multiple Codex chats from a TODO list view, which will be executed in parallel.
It will become very instrumental in the upcoming Codex Superapp, where you will be able to… pic.twitter.com/YFquUwwRMxOpenAI is working on a new experimental feature for Codex called Scratchpad. Users will be able to start multiple Codex chats from a TODO list view, which will be executed in parallel. It will become very instrumental in the upcoming Codex Superapp, where you will be able to
-
Tools and Harnessing Drive Recent AI Progress Over Scale
By
–
nope, tools/harnessing is what is accounting for most of the recent progress. not scale. no way could you get a pure LLM to do what Claude Code does. that’s why they used the other stuff, too see eg https://
open.substack.com/pub/garymarcus
/p/how-o3-and-grok-4-accidentally-vindicated?r=8tdk6&utm_medium=ios
… -
API Provider Features and Prompt Caching in Deep Agents
By
–
this is a good point around taking advantage of model/api provider features i agree that prompt caching is great! we make sure to use it in deepagents! but that alone doesnt lock in – you can switch pretty easily, just a bit more costly things like encrypted or serverside
-
Composer 2 Usage Doubled in New Interface Update
By
–
This weekend we're doubling Composer 2 usage in our new interface.
— Cursor (@cursor_ai) 11 avril 2026
Open Agents Window, pick Composer 2, and start building! No hourly limits.https://t.co/Z5GHz7kcG1This weekend we're doubling Composer 2 usage in our new interface. Open Agents Window, pick Composer 2, and start building! No hourly limits.
-
Prompt Caching vs Vendor Lock-in in AI Agents
By
–
i agree that prompt caching is great! we make sure to use it in deepagents! but that alone doesnt lock in – you can switch pretty easily, just a bit more costly things like encrypted or serverside compaction make it harder to do so. blackbox long term memory makes it impossible
-
Code and Prompts: Testing and Evaluation Strategies
By
–
Combination of code and prompts. Need tests and evals
-
Claude for Word Beta: AI Integration Nails Microsoft Office Productivity
By
–
🚨 Anthropic just made Microsoft Word cool again.
— Charly Wargnier (@DataChaz) 11 avril 2026
“Claude for Word” is officially in beta.
Now you get Claude right in your sidebar. NO MORE TAB SWITCHING 🤯
I think they completely nailed the execution on this one.
The top 4 mind-blowing features:
1. Cross-app brain
→ It… pic.twitter.com/MuWccy1TDv🚨 Anthropic just made Microsoft Word cool again. “Claude for Word” is officially in beta. Now you get Claude right in your sidebar. NO MORE TAB SWITCHING 🤯 I think they completely nailed the execution on this one. The top 4 mind-blowing features: 1. Cross-app brain → It can analyze your open Microsoft Word, Microsoft Excel, and Microsoft PowerPoint files at the same time (!) 2. Native tracked changes → It suggests edits like a human coworker, it doesn’t just overwrite your work. 3. Formatting that actually works → It respects your H1s, H2s, and even complex numbering. 4. Team skills → You can turn your best prompts into reusable tools for your whole team. IMO this is the most practical AI integration I’ve seen all year. Who’s testing this today?