Toots for simon@simonwillison.net account

Written by Simon Willison on 2025-01-22 at 03:50

Here's a fun clue for people who are obsessing over how much time has passed since S1 (Milchick said 5 months, but why would we believe anything any Lumon employee says about anything?)

Someone on MetaFilter just pointed out that the only solid evidence we have that any time has passed at all is that it has to cover the production time needed to produce a claymation corporate video https://fanfare.metafilter.com/25332/Severance-Hello-Ms-Cobel#348398

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-22 at 03:41

New release of the llm-gemini plugin - a couple of new experimental models, but the really fun feature is "-o google_search 1" to enable a feature where models can run additional Google searches as part of responding to a prompt.

https://github.com/simonw/llm-gemini/releases/tag/0.9

=> View attached media

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-22 at 02:11

I just published the videos from our recent Datasette Public Office Hours, with six different demos from six different presenters of new projects built on top of Datasette and LLM https://simonwillison.net/2025/Jan/22/office-hours-demos/

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-21 at 22:07

If you have a browser that supports WebGPU (like Google Chrome) you try out the DeepSeek-R1 model based on Qwen2.5-Math-1.5B directly in your browser!

It's a 1.28GB page load: https://huggingface.co/spaces/webml-community/deepseek-r1-webgpu

=> View attached media

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-21 at 18:03

Updated my post with some alternative ways to access DeepSeek R1: https://simonwillison.net/2025/Jan/20/deepseek-r1/#other-ways-to-try-deepseek-r1

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-20 at 15:22

DeepSeek released a whole family of inference-scaling / "reasoning" models today, including distilled variants based on Llama and Qwen

Here are my notes on the new models, plus how I ran DeepSeek-R1-Distill-Llama-8B on my Mac using Ollama and LLM

https://simonwillison.net/2025/Jan/20/deepseek-r1/

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-19 at 01:40

Alternative URL for if my site isn't responding https://github.com/simonw/til/blob/main/tiktok/download-all-videos.md

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-19 at 01:12

Wrote up a gnarly sequence of hacks for downloading every video in a TikTok account, just in case anyone needs such a thing in the next ~24 hours

uv, yt-dlp, JS in the Firefox DevTools and optionally throws in mlx-whisper for generating transcripts too https://til.simonwillison.net/tiktok/download-all-videos

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-17 at 22:05

Today's Datasette Public Office Hours session is now broadcasting live on YouTube! https://www.youtube.com/watch?v=Fq_oggE712Q

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-16 at 18:42

We're running a Datasette Public Office Hours event on Discord tomorrow (Friday 2pm PT) with lightning demos from community members of their Datasette and LLM-adjacent projects

We still have some space left, so if you have something you'd like to demo please sign up here! https://simonwillison.net/2025/Jan/16/datasette-public-office-hours/

=> View attached media

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-15 at 01:42

My notes on the new OpenAI ChatGPT scheduled tasks feature, including a copy of the new tool instructions from a leaked system prompt https://simonwillison.net/2025/Jan/15/chatgpt-tasks/

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-13 at 21:44

My notes so far on Codestral 25.01 - the new code-focused API-only LLM released today by Mistral https://simonwillison.net/2025/Jan/13/codestral-2501/

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-12 at 01:54

Extraordinary article on the energy usage of generative AI from BloombergNEF founder Michael Liebreich - absolutely worth spending some time with this: https://about.bnef.com/blog/liebreich-generative-ai-the-power-and-the-glory/

I wrote up some of my own notes on the article here: https://simonwillison.net/2025/Jan/12/generative-ai-the-power-and-the-glory/

=> More informations about this toot | View the thread

Shared by Simon Willison on 2025-01-10 at 21:49 (original by Scott Klein)

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-10 at 01:45

I am SO MUCH more comfortable writing about things that have actually happened as opposed to pontificating on whatever the next few months and years might bring https://simonwillison.net/2024/Dec/31/llms-in-2024/

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-10 at 01:44

Risky post... here are my 1, 3 and 6 year predictions for LLMs/AI, expanded from my appearance on the Oxide and Friends podcast

(My confidence in my ability to predict the future is extremely low)

https://simonwillison.net/2025/Jan/10/ai-predictions/

=> More informations about this toot | View the thread

Shared by Simon Willison on 2025-01-09 at 02:48 (original by Jeff Triplett)

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-07 at 00:52

I'll be participating in this live audio podcast recording in ten minutes time on the Oxide Discord https://mastodon.social/@bcantrill/113760041070406490

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-04 at 23:57

First weeknotes of 2025 https://simonwillison.net/2025/Jan/4/weeknotes/

=> More informations about this toot | View the thread

Written by Simon Willison on 2025-01-03 at 17:51

(To be fair here, I produce such a high volume of content that it's understandable when people evaluate my work based on a single post without taking the breadth of my coverage into account)

=> More informations about this toot | View the thread

=> This profile without reblog | Go to simon@simonwillison.net account