Simon
@tokumin
Followers
6K
Following
9K
Media
200
Statuses
7K
@NotebookLM lead @GoogleLabs đź•’ NotebookLM Audio Overviews, Gemini & PaLM2 post-training, AI Studio, YouTube, Discover, Search, Android. More LaMDA moments
latent space
Joined February 2007
RT @lintool: At @yupp_ai we add new models as soon as they drop! @sundarpichai tweeted about Gemini 2.5 Flash-Lite yesterday. Soon after, i….
0
8
0
the new Gemini-2.5-Flash-Lite is as close as we have gotten to @karpathy's cognitive core. FL is a really capable model and bending my mind about where and for what AI can be used for in products and the leverage that having a whole family of models gives you at the small end of.
The race for LLM "cognitive core" - a few billion param model that maximally sacrifices encyclopedic knowledge for capability. It lives always-on and by default on every computer as the kernel of LLM personal computing. Its features are slowly crystalizing:. - Natively multimodal.
13
33
579
really interesting trend in personal projects that literally can never be made by a company for all kinds of reasons.
I’ve started taking pictures of nearly every meal I eat to feed into a multimodal AI doctor that has context on my wearable, how I’m feeling, blood work, meds/supplements, genome variants interpreted with a reasoning model, etc. It’s a very hacky prototype now but I’m surprised.
0
1
5
what's the best way to get a machine readable version of my X timeline? I paid $200/mođź’€ to see if the basic tier of the X API has it (the docs are impenetrable) but still getting a 403 from the timeline reading API. do people use something browser based like playwright mcp?.
3
0
7
agent demos are hard because almost by definition they are like watching someone else doing time-consuming boring knowledge work. memory and p13n demos suffer from a related problem. you just have to experience them yourself and make your own mind up.
1
0
15
since Gemini CLI is OSS, you can repomix it and see that it's actually architected as a MCP supporting headless Gemini agent and CLI is only one of many possible heads. you can fork it and make your own high quality Gemini agent.
github.com
An open-source AI agent that brings the power of Gemini directly into your terminal. - google-gemini/gemini-cli
1
3
19
you can tell pretty quickly which models are born of a purely scientific process and which incorporate artistry. Kimi has some extremely elegant, artistic touches that make using it a very delightful experience. bravo to the RL team.
0
1
18
oh great.
japantoday.com
In July 1999, manga author Ryo Tatsuki published "Watashi ga Mita Mirai" (The Future I Saw) which was based on her own dream journal she had been keeping for years. In the manga, she explains how the...
There's a new interstellar comet in the neighborhood! . Known as 3I/ATLAS, this comet poses no threat to Earth – but it does provide a rare opportunity to study an object that originated outside of our solar system:
0
0
5
I think OSS LLMs are cool, but I've spoken to startups and companies the world over. <2% are using OSS models because they are a pain to run & maintain, let alone scale. >98% are using API models like Flash or *mini and enjoying the ride.
0
0
11
it's one year later. you've still not read situational awareness. make this week the week you read it.
situational-awareness.ai
Leopold Aschenbrenner, June 2024 You can see the future first in San Francisco. Over the past year, the talk of the town has shifted from $10 billion compute clusters to $100 billion clusters to...
1
1
24
every now and then I check-in on the WizardLM team. they were so ahead of the curve with their work on Evol.
Imagine where Microsoft would be now if they hadn't destroyed the amazing WizardLM team, causing them to leave and go to Tencent…. …where they're now releasing models far beyond anything Microsoft has created.
0
0
8
RT @trondw: I'm super proud of the team behind Doppl, our newest experiment from Google Labs! Available now on Android and iOS:. https://t….
play.google.com
See any style, try on any look.
0
5
0
NotebookLM now available directly on the shelf of new Lenovo Chromebook Plus 14!.
As of today, the new @Lenovo Chromebook Plus 14 is available and comes with a whole suite of new AI features, including access to @NotebookLM directly from the shelf. From web, to mobile, now to direct on desktops— where should we go next? 🗺️
0
0
16
"Simon, you are asking a machine for analysis on a day made for synthesis; stop querying the data of your past and instead, go feel the verdict of the midday sun on this peak summer Sunday—its warmth is the only answer you need right now." => nods head.
0
0
5
Building Software is being automated but Building Products is an ongoing, mostly manual RLHF process searching for fitness in a changing landscape. The team that automates the full build, measure, learn loop of product development will unlock significant velocity. spending a.
product development needs an AI native overhaul. a preview of what's to come.
0
0
5