Jacky Casas

Co-founder at PlaynVoice. PhD in CS.

[Jacky Casas @jackycasas_]

@romaindewolff Crazy story! Thanks for sharing, we have no spending limit set yet, will do asap

[Romain de Wolff @romaindewolff]

I think it's the first time i am mentioned in a podcast from another founder ❤️

[Jacky Casas @jackycasas_]

@romaindewolff Does it mean I have to start building in public as well?

[kepano @kepano]

Defuddle now has a website!

This means you can use Defuddle anywhere to get the main content of a page in Markdown format.

You can simply add "defuddle.md" before any URL, use it via curl, Skills, CLI, or add it to your app via NPM.
https://t.co/AXwXCvRtgI

[SwissDevJobs.ch @SwissDevJobs]

👨‍💻 Discover PlaynVoice by Jacky Casas!

PlaynVoice is an AI-powered scribe helping psychologists and therapists in Switzerland save over an hour a day by automatically generating consultation notes.

Platform: https://t.co/tfHoitXrmB

#SwissDevJobs

[Jacky Casas @jackycasas_]

@Vapi_AI You guys rock!

[Cognition @cognition]

Today we're excited to introduce Devin, the first AI software engineer.

Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork.

Devin is an autonomous agent that solves engineering tasks through the use of its own shell, code editor, and web browser.

When evaluated on the SWE-Bench benchmark, which asks an AI to resolve GitHub issues found in real-world open-source projects, Devin correctly resolves 13.86% of the issues unassisted, far exceeding the previous state-of-the-art model performance of 1.96% unassisted and 4.80% assisted.

Check out what Devin can do in the thread below.

[Austin Vance @austinbv]

I have had the privilege to use LangSmith and from @langchain over the last month or so and it's downright amazing...

With LangSmith now going GA I wanted to share a few things that I have absolutely loved about it.

And shoot me a DM if you want to know more

🧵 👇

[Sam Altman @sama]

the mission continues

[LangChain @LangChain]

⭐️ Multi-Vector Retriever for RAG on tables, text, and images ⭐️

Seamless question-answering across diverse data types (images, text, tables) is one of the holy grails of RAG.

We’re releasing three new cookbooks that showcase the multi-vector retriever to tackle this challenge.

We released the multi-vector retriever back in August w/ a simple idea:

1/ embed a doc reference (e.g., summary) that is optimized for search

2/ but retrieve the raw doc (table, text, image) to give complete context for LLM answer synthesis

Using @UnstructuredIO to parse images, text, and tables (e.g., from pdfs) ...

.. the multi-vector retriever enables RAG on semi-structured data w/ table summaries -
https://t.co/Vmxcifa81A

.. we extend the idea to images, using LLaVA-7b (c/o @imhaotian) to produce image summaries -
https://t.co/PufDyCGM9J

... and this full RAG pipeline can be run laptop w/ llama.cpp c/o @ggerganov, @ollama_ai, @nomic_ai embeddings, and @trychroma:
https://t.co/yXxlxGkjbq

Blog:
https://t.co/U5tymTXdYY

[Valtteri Karesto @valtterikaresto]

Finite state machines and LLMs? Anyone toying with these?

[Product Hunt 😸 @ProductHunt]

We have entered a new era of ChatGPT.

In the span of a few days, OpenAI has announced not one, but FOUR new updates for the chatbot.

Here's a recap ⬇️

[Andrej Karpathy @karpathy]

With many 🧩 dropping recently, a more complete picture is emerging of LLMs not as a chatbot, but the kernel process of a new Operating System. E.g. today it orchestrates:

- Input & Output across modalities (text, audio, vision)
- Code interpreter, ability to write & run programs
- Browser / internet access
- Embeddings database for files and internal memory storage & retrieval

A lot of computing concepts carry over. Currently we have single-threaded execution running at ~10Hz (tok/s) and enjoy looking at the assembly-level execution traces stream by. Concepts from computer security carry over, with attacks, defenses and emerging vulnerabilities.

I also like the nearest neighbor analogy of "Operating System" because the industry is starting to shape up similar:
Windows, OS X, and Linux <-> GPT, PaLM, Claude, and Llama/Mistral(?:)).
An OS comes with default apps but has an app store.
Most apps can be adapted to multiple platforms.

TLDR looking at LLMs as chatbots is the same as looking at early computers as calculators. We're seeing an emergence of a whole new computing paradigm, and it is very early.

[Jacky Casas @jackycasas_]

@neon_accountapp Very good, thanks for the reply ;)

[Yann LeCun @ylecun]

This is huge: Llama-v2 is open source, with a license that authorizes commercial use!

This is going to change the landscape of the LLM market.
Llama-v2 is available on Microsoft Azure and will be available on AWS, Hugging Face and other providers

Pretrained and fine-tuned models are available with 7B, 13B and 70B parameters.

Llama-2 website: https://t.co/PKrrXgHdem
Llama-2 paper: https://t.co/aINNrXNhMb

A number of personalities from industry and academia have endorsed our open source approach: https://t.co/N7HwgW9Suh

[Marco Mascorro @ ICML @Mascobot]

✨NEW LAUNCH! LLaMA2 chat API &amp; open-source playground💫:

We're releasing tools that make it easy to test @meta's latest LLM &amp; add it to your own app with @replicatehq.

Playground: https://t.co/YRxDyl5fVW
Live chat API here: https://t.co/TFUOsy44oT

Repos &amp; instructions below:

[Andrej Karpathy @karpathy]

Huge day indeed for AI and LLMs, congrats to Meta 👏
This is now the most capable LLM available directly as weights to anyone from researchers to companies.

The models look quite strong, e.g. Table 4 in the paper: MMLU is good to look at, the 70B model is just below GPT-3.5. But HumanEval (bad misnomer) shows coding capability is quite a bit lower (48.1 vs 29.9).

[Steven Tey @steventey]

▲ @Vercel just launched the AI SDK.

You can now build AI-powered apps with streaming – starting with one simple command: `npm i ai`

Here are some examples👇

[BigCode @BigCodeProject]

Introducing: 💫StarCoder

StarCoder is a 15B LLM for code with 8k context and trained only on permissive data in 80+ programming languages. It can be prompted to reach 40% pass@1 on HumanEval and act as a Tech Assistant.

Try it here: https://t.co/4XJ0tn4K1m

Release thread🧵