DPS Build
记录了一下最近使用 Carrd 连接自定义 API 的经历。 Carrd 是由 AJ 一个人创建的建站服务，2016年发布之后，两年就达到了一百万美金的 ARR。 Carrd 非常容易上手，这次我们尝试使用一个 custom form 的功能，遇到了一些困难。中间遇到了一些困难。开了 ticket 之后，AJ 一直跟进，工作日基本上一小时内就有回复，休息日也会在24小时内回复。非常佩服他的响应速度！ https://letters.acacess.com/carrd-with-apis/
惊叹 AJ 的效率和用户关怀。
前几天和他反映，custom form 的 timeout 太短，无法满足需求。他回复说，会考虑延长 timeout。
昨天他回复说，已经延长了。
前几天和他反映，custom form 的 timeout 太短，无法满足需求。他回复说，会考虑延长 timeout。
昨天他回复说，已经延长了。
这个 thread 太好玩了：改编了各大网站的 slogan，比如 Webflow — 祝你能够猜出我们的价格
https://twitter.com/jovvvian/status/1643663422863863809
https://twitter.com/jovvvian/status/1643663422863863809
X (formerly Twitter)
Jovian Gautama 劉恆原 (@jovvvian) on X
Tech companies honest headlines: a thread
1. Calendly
1. Calendly
iPhone 14 Pro 上的 neural engine (ANE) 比 MacBook Pro M1 Max 上的还要qiang jing
https://twitter.com/bjnortier/status/1643966251188748288
https://twitter.com/bjnortier/status/1643966251188748288
Twitter
What surprised me is that Whisper.cpp+CoreML is faster on my iPhone 14 Pro A16 Bionic than my MacBook Pro M1 Max, 713 vs 773ms for one sample 🤯 I suspect it's because the M1 has the same Apple Neural Engine as the A14 Bionic. #OpenAI #Whisper #AI
https:…
https:…
两个新的薪资透明网站：
1. https://www.comprehensive.io
2. https://www.levels.fyi/jobs
https://blog.pragmaticengineer.com/the-scoop-salary-transparency/
1. https://www.comprehensive.io
2. https://www.levels.fyi/jobs
https://blog.pragmaticengineer.com/the-scoop-salary-transparency/
www.comprehensive.io
Compensation Management Platform | Comprehensive
Compensation management software that streamlines pay review cycles, salary benchmarking, and total rewards statements — without a thousand spreadsheets.
macOS 里藏着 Bitcoin 的白皮书，在 Terminal 里输入以下命令即可打开：
open /System/Library/Image\ Capture/Devices/VirtualScanner.app/Contents/Resources/simpledoc.pdfhttps://waxy.org/2023/04/the-bitcoin-whitepaper-is-hidden-in-every-modern-copy-of-macos/
Waxy.org
The Bitcoin Whitepaper Is Hidden in Every Modern Copy of macOS - Waxy.org
I just discovered that every copy of macOS ships with a hidden PDF of Satoshi Nakamoto's Bitcoin whitepaper. But why?
👍1
Online services often exhibit data locality, with users frequently accessing popular or trending content. Cache systems take advantage of this behavior by storing commonly accessed data, which in turn reduces data retrieval time, improves response times, and eases the burden on backend servers. Traditional cache systems typically utilize an exact match between a new query and a cached query to determine if the requested content is available in the cache before fetching the data.
However, using an exact match approach for LLM caches is less effective due to the complexity and variability of LLM queries, resulting in a low cache hit rate. To address this issue, GPTCache adopt alternative strategies like semantic caching. Semantic caching identifies and stores similar or related queries, thereby increasing cache hit probability and enhancing overall caching efficiency.
GPTCache employs embedding algorithms to convert queries into embeddings and uses a vector store for similarity search on these embeddings. This process allows GPTCache to identify and retrieve similar or related queries from the cache storage, as illustrated in the Modules section.
https://github.com/zilliztech/gptcache
However, using an exact match approach for LLM caches is less effective due to the complexity and variability of LLM queries, resulting in a low cache hit rate. To address this issue, GPTCache adopt alternative strategies like semantic caching. Semantic caching identifies and stores similar or related queries, thereby increasing cache hit probability and enhancing overall caching efficiency.
GPTCache employs embedding algorithms to convert queries into embeddings and uses a vector store for similarity search on these embeddings. This process allows GPTCache to identify and retrieve similar or related queries from the cache storage, as illustrated in the Modules section.
https://github.com/zilliztech/gptcache
GitHub
GitHub - zilliztech/GPTCache: Semantic cache for LLMs. Fully integrated with LangChain and llama_index.
Semantic cache for LLMs. Fully integrated with LangChain and llama_index. - GitHub - zilliztech/GPTCache: Semantic cache for LLMs. Fully integrated with LangChain and llama_index.
前几天和朋友聊到微软推出的 visualchat，一看，好家伙，22个模型串起来，至少要四张卡才能跑起来。
https://github.com/microsoft/visual-chatgpt#gpu-memory-usage
https://github.com/microsoft/visual-chatgpt#gpu-memory-usage
GitHub
GitHub - chenfei-wu/TaskMatrix
Contribute to chenfei-wu/TaskMatrix development by creating an account on GitHub.
DPS Build
OpenAI 的 Lilian Li 专门介绍了 prompt engineering https://t.me/tms_ur_way/2655
Chip Huyen 写了一篇非常详细的 当下 LLM 产品化总结：
1. 目前的 prompt engineering 大大降低了开发难度，同时也提高了维护成本；
2. 尽管直接调用各种 LLM API 可能会产生天价费用，但是比起自己从零训练模型还是便宜；
3. LLM 输出可能不稳定，需要通过各种方法来降低这种风险，比如在 prompt 中加入更多的例子；
4. LLM 的 latency 可能也是一个潜在问题
5. 总之，这一领域发展得非常快，也许这些经验中的很大一部分在三个月后就没有参考意义了。
https://readwise.io/reader/shared/01gxwbcx0r0bf7tnbcn1nh1nw4
1. 目前的 prompt engineering 大大降低了开发难度，同时也提高了维护成本；
2. 尽管直接调用各种 LLM API 可能会产生天价费用，但是比起自己从零训练模型还是便宜；
3. LLM 输出可能不稳定，需要通过各种方法来降低这种风险，比如在 prompt 中加入更多的例子；
4. LLM 的 latency 可能也是一个潜在问题
5. 总之，这一领域发展得非常快，也许这些经验中的很大一部分在三个月后就没有参考意义了。
https://readwise.io/reader/shared/01gxwbcx0r0bf7tnbcn1nh1nw4
huyenchip.com | highlighted via Readwise
Building LLM applications for production | annotated by
A question that I’ve been asked a lot recently is how large language models (LLMs) will change MLOps workflows. After working with several companies who have...
❤1👍1
经朋友推荐，看到这个域名搜索引擎，非常不错。
https://domainr.com/
比如你想买一个 telegram 的域名，它会把相近的候选项自动列举出来。比如 telegr.am, Tele.gram 等组合
https://domainr.com/
比如你想买一个 telegram 的域名，它会把相近的候选项自动列举出来。比如 telegr.am, Tele.gram 等组合
👍1
How Replit trains Large Language Models (LLMs) using Databricks, Hugging Face, and MosaicML
https://blog.replit.com/llm-training
https://blog.replit.com/llm-training
Replit Blog
Replit — How to train your own Large Language Models
Learn how Replit trains Large Language Models (LLMs) using Databricks, Hugging Face, and MosaicML
Introduction
Large Language Models, like OpenAI's GPT-4 or Google's PaLM, have taken the world of artificial intelligence by storm. Yet most companies don't…
Introduction
Large Language Models, like OpenAI's GPT-4 or Google's PaLM, have taken the world of artificial intelligence by storm. Yet most companies don't…