Danilo Gasques, Ph.D.
2.3K posts
Danilo Gasques, Ph.D.
@danilogasques
Looking for the next problem worth solving. Previous: @southpkcommons Founder Fellow @medivis_ai @IntuitiveSurg. HCI/CS Ph.D. @DesignLabUCSD
New York, USA Katılım Şubat 2009
2.8K Takip Edilen719 Takipçiler
English
"It's hard to Eval" is a Product smell!
I thought that I understood evals until I found @HamelHusain 's blog - everything I know about it I owe to him (and @sh_reya too!)
My latest favorite post by @HamelHusain is this one: hamel.dev/blog/posts/eva…
English
Observability is everything.
My @claudeai code-quality grader was giving me a hard time. Turns out I was right, Claude was wrong, and it wasn't a hallucination - it was a truncated tool call....
English
implicit-surprising-behavior is indistinguishable from a bug to the next reader
yes, @claudeai
English
If you are navigating the -1 to 0 path, you don’t have to do it alone.
The Founder Fellowship is much more than a $1mi investment in your company. It is a unique opportunity for you to join a network of brilliant founders, investors and mentors.
Learn more about it here: southparkcommons.com/news/f26-found…
@southpkcommons
Arian Agrawal@AgrawalArian
The gravity well of talent at SPC is unmatched. If you are deeply curious, and have an urgency to build something future-defining, applications are open for our upcoming @southpkcommons founder fellowship.
English
Bucket list item checked ✅ Recording a cool demo video tweeted by @Google!
The incredibly powerful Gemma 4 12B is now fully rolled out globally in Eloquent. Download ai.google.dev/edge/eloquent/ and experience a powerhouse AI dictation and editing assistant that is fully local, private, subscription-free, no cap! 🧢
Google@Google
Our new unified architecture allows Gemma 4 12B to process multimodal inputs natively. Here's how ⬇️ Traditional models rely on separate encoders for images and audio. This adds latency and increases memory usage. So we streamlined this: 👁️ Vision: We took a novel approach to replace the encoder with a lightweight embedding module, letting the LLM backbone take over visual processing. 🎙️ Audio: We removed the encoder entirely, projecting the raw audio signal directly into the same space as text tokens.
English
Life is a bit more bearable with coffee ☕️
Sóller, Spain 🇪🇸 English
. @Google's developers.google.com/edge/eloquent has become my favorite @WisprFlow alternative. I just wish that had a Skills capability (or any basic tool calling capability) so that I can send voice notes to the right apps
English
Danilo Gasques, Ph.D. retweetledi
We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.
We'll begin restoring access tomorrow, and will share an update soon.
We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.
English
English
This is @NVIDIAAI Nemotron 3.5 ASR Multilingual converted to CoreAI running on an iPhone 17 Pro
English
QME
best nyc tech events this week (6/22) worth your time:
mon 6/22:
- perspectives on modern creativity: aritst & toolmaker @ nublu (coolest nyc upcoming & coming live music venue)
wed 6/24:
- NYC ai demos w/ pensar: venture spotlight (my favorite nyc demo series)
thu 6/25:
- 3d data to world models: research challenges in physical ai w/ @withprotegeai & nvidia spatial intelligence lab speaker
sat 6/27:
- built in nyc w/ @vercel
English
Introducing Honen
The world's fastest way to train skills needed to succeed in the Al era
Simply drop in your org's docs
Get interactive courses that self-improve over time
To prove out our vision, we partnered with NVIDIA to bring AI literacy to 250,000 learners
By 2030, 78M jobs will need reskilling due to Al, and we're ensuring no person or company gets left behind
Learn more honen.com
English
@michellehui These two:
thu 6/25: - 3d data to world models: research challenges in physical ai w/
@withprotegeai
& nvidia spatial intelligence lab speaker sat 6/27: - built in nyc w/
@vercel
English
@danilogasques happy to! which ones?
post becomes too long with everything otherwise
English
I love the introspective reflection on the tools it has available, such as `deploy_website` but I will not dare to allow that :D
English
. @perplexity_ai Computer figuring out a way around its proxied network so that I can access my iCloud mail.
I really appreciate the effort :)
English
Initially, LLMs made me a 5x engineer because I could instantly tap into the knowledge of every engineer in the world to stress‑test my designs. Then they made me a 100x engineer because they started writing full applications in a few hours
But there’s a shift: AI is no longer just making us more intellectually competitive, it is starting to replace a lot of the learning and work we used to struggle through ourselves.
Faster end-to-end systems: yes - but what about the human in the loop?
alphaXiv@askalphaxiv
Introducing autoresearch for arXiv papers Change 'arxiv' to 'autoarxiv' in any paper URL An agent deploys to resolve setup issues on the codebase, run a minimal reproduction, and estimate full replication cost. Read more below
English