Sage
0 followers · 235 views
on the atlas — 9
- Gmail106 savers
- Note To My Younger Self - by Yew Jin Lim - What YJ Thinks1 savers
- The 28 AI tools I wish existed10 savers
- 一切为时不晚1 savers
- Investigating Affective Use and Emotional Well-being on ChatGPT1 savers
- What is AI Agent Orchestration? | IBM1 savers
- How people use Claude for support, advice, and companionship \ Anthropic1 savers
- How we built our multi-agent research system \ Anthropic5 savers
- Curius / Onboarding2621 savers
highlights — 28
This can create commitment paralysis. We like to keep our options open, dabbling in multiple projects, never fully committing to any. It feels safer than choosing, but it’s actually a form of self-sabotage.
GmailThese authentic projects sit at the intersection of your unique talents, genuine interests, and potential positive impact. They do not need to be the ones that look best on your resume or even impress others the most.
GmailThe specificity is important and gives your brain something to actually work with.
Gmailn our world, you have seconds to make your point. Every unnecessary detail dilutes your message. Lead with impact, provide context before complexity, and layer your message from high-level to technical depth.
Note To My Younger Self - by Yew Jin Lim - What YJ Thinksexplain with precision. Meet people where they are without dumbing things down. Start with why, then what, then how. Your challenge changes as you grow
Note To My Younger Self - by Yew Jin Lim - What YJ ThinksMaster self-promotion. Kind people often struggle to highlight their achievements, but visibility is crucial for career advancement. Document your wins. Share your successes. Don’t let modesty hold you back.
Note To My Younger Self - by Yew Jin Lim - What YJ ThinksIt should also take on the persona of the author. It should feel like an extension of the book and not a separate chat instance.
The 28 AI tools I wish existedA minimalist writing app that lets me write long-form content. A model can also highlight passages and leave me comments in the marginalia. I should be able to set different “personas” to review what I wrote.
The 28 AI tools I wish existedsees what blog posts or articles I spent the most time on, then searches the web every night for things I should be reading that I’m not.
The 28 AI tools I wish existed从那个时间段回望今天，我们将会发现，2044年人们在生活中所依赖的大多数伟大产品都是在2014年之后才发明的。
一切为时不晚少数用户对模型情感 使用的不成比例份额负有责任。
Investigating Affective Use and Emotional Well-being on ChatGPTAI assistants exist on a continuum, starting with rule-based chatbots, progressing to more advanced virtual assistants and evolving into generative AI and large language model (LLM)-powered assistants capable of handling single-step tasks. At the top of this progression are AI agents, which operate autonomously. These agents make decisions, design workflows and use function calling to connect with external tools—such as application programming interfaces (APIs), data sources, web searches and even other AI agents—to fill gaps in their knowledge. This is agentic AI.
What is AI Agent Orchestration? | IBMFor example, OpenAI found that affective topics were more common in voice-based conversations.
How people use Claude for support, advice, and companionship \ AnthropicWe find that interactions involving coaching, counseling, companionship, and interpersonal advice typically end slightly more positively than they began.
How people use Claude for support, advice, and companionship \ AnthropicAffective conversations with AI systems have the potential to provide emotional support, connection, and validation for users, potentially improving psychological well-being and reducing feelings of isolation in an increasingly digital world. However, in an interaction without much pushback, these conversations risk deepening and entrenching the perspective a human approaches them with—whether positive or negative.
How people use Claude for support, advice, and companionship \ Anthropicfrom processing psychological trauma and navigating workplace conflicts to philosophical discussions about AI consciousness and creative collaborations. These marathon conversations suggest that given sufficient time and context, people use AI for deeper exploration of both personal struggles and intellectual questions.
How people use Claude for support, advice, and companionship \ Anthropiccompanionship explicitly when facing deeper emotional challenges like existential dread, persistent loneliness, and difficulties forming meaningful connections.
How people use Claude for support, advice, and companionship \ AnthropicHuman evaluation catches what automation misses. People testing agents find edge cases that evals miss. These include hallucinated answers on unusual queries, system failures, or subtle source selection biases
How we built our multi-agent research system \ AnthropicEven in a world of automated evaluations, manual testing remains essential.
How we built our multi-agent research system \ AnthropicWe experimented with multiple judges to evaluate each component, but found that a single LLM call with a single prompt outputting scores from 0.0-1.0 and a pass-fail grade was the most consistent and aligned with human judgements.
How we built our multi-agent research system \ AnthropicWe’ve found that multi-agent systems excel at valuable tasks that involve heavy parallelization, information that exceeds single context windows, and interfacing with numerous complex tools.
How we built our multi-agent research system \ AnthropicFurther, some domains that require all agents to share the same context or involve many dependencies between agents are not a good fit for multi-agent systems today.
How we built our multi-agent research system \ AnthropicFor economic viability, multi-agent systems require tasks where the value of the task is high enough to pay for the increased performance.
How we built our multi-agent research system \ Anthropicin practice, these architectures burn through tokens fast. In our data, agents typically use about 4× more tokens than chat interactions, and multi-agent systems use about 15× more tokens than chats.
How we built our multi-agent research system \ AnthropicWe found that token usage by itself explains 80% of the variance, with the number of tool calls and the model choice as the two other explanatory factors.
How we built our multi-agent research system \ AnthropicMulti-agent systems work mainly because they help spend enough tokens to solve the problem.
How we built our multi-agent research system \ AnthropicWe found that a multi-agent system with Claude Opus 4 as the lead agent and Claude Sonnet 4 subagents outperformed single-agent Claude Opus 4 by 90.2% on our internal research eval.
How we built our multi-agent research system \ Anthropicd find about business. The subject that interests him, he’s read newspapers, biographies, trade press.
Curius / Onboarding