Sabitlenmiş Tweet
Ingrid
252 posts
Ingrid
@ingowoo
pizza and hackathons | ex @google @tiktok_us @uber
Katılım Ekim 2022
588 Takip Edilen631 Takipçiler
Ingrid retweetledi
卧槽
Claude 动手抄 OpenAI 老家了
一键把你在 ChatGPT 攒的记忆全搬走
Anthropic 上线了一个记忆迁移工具（Memory Import），让你把 ChatGPT、Gemini 等 AI 助手里积累的个人偏好和上下文，一键导入到 Claude 的记忆系统里。
众所周知 ChatGPT 的一大核心竞争力就是其 记忆系统 很完善...
你用的越久，GPT越懂你，你越离不开。
Claude这个工具简直是釜底抽薪...
Claude 官方这个工具的最大优势是：直接写入 Claude 记忆系统，不需要绕路，不需要装插件，三步完成。
中文
Perfectly captures what i've been feeling talking to startups in 2025: a16z.com/the-palantiriz…
Every company with an AI product (Posthog, Notion, Hubspot, ...) has realized the difficulties of "last mile" deployment, launched a workflow orchestrator, and is now trying enterprise expansion thru palantirization. Risks turning startups into glorified consultancies, esp if they lack strong primitives and big buck customers.
English
Ingrid retweetledi
Google Meet show how late people will be based on meeting history
English
Students @UWaterloo or @UofT, my friends at @raycasterai (YC '24) are hosting a chat about AI for Regulatory Sciences on 1/6~! Check them out luma.com/ejotlqyd
English
Please don’t do temperature < 1. It‘s not worth it, esp. if you‘re building an agent.
Reduces:
- function calling quality
- thought quality
- token efficiency
- error recovery in agents
- markdown consistency
… and it also does not give you more factual results.
If you need more predictability, design a good system instruction. Temp < 1 is a relic of Q&A apps.
Most $$M spend customers I have worked with: temp=0.1 -> temp=1 -> problem solved.
English
@ishaansehgal you're comparing hiring for an entry-level job vs 10x engineer
English
what founders are hiring for now:
- can you ship with cursor
- have you built with claude
- do you actually code or just prompt
what they're NOT hiring for:
- read every AI paper
- can explain transformers
- went to that AI conference
talk less, build more
English
@Pri_promo Whenever I see one of these, I assume the company is pre-product
English
good on humanity to spend centuries weaving social reward systems that nudge us toward thought. So now we effectively subsidize the metabolic cost of thinking.
signüll@signulll
most ppl don’t really think. they rearrange cached thoughts until they feel smart. true thinking is rare because it’s metabolically expensive (just like thinking models are computationally expensive).
English
Played around with @LeRobotHF and SO-101 and amazed at how accessible training VLAs has become. What an interesting way to exhibit telepresence! Wish id tried this ~
Yu Xiang@YuXiang_IRVL
Our students are presenting their robotics projects this week! I’ll be sharing some of the coolest ones over the next few days. The first one is a dancing SO-101😄Turn on your volume, there’s music! #Robotics
English
@mbrendan1 Hi @mbrendan1 a colleague of mine told me about this program but it seems I’m a tad late. Is there any way to get involved remotely?
English
Share your project proposal by Dec 6th.
Apply: cosmosinst.typeform.com/fellowship-2026
English
Cosmos Fellowship applications are open: your chance to join the @UniofOxford's Human-Centered AI Lab with @PhilippKoralus.
We’re looking for individuals with the potential for world-class AI expertise and deep philosophic insight 👇
English