mott

Resend just crossed 92k customers. 12 months ago: 9k 6 months ago: 24k 3 months ago: 49k today: 92k whatever you're building, the slow start isn't failure. it's where most people quit. don't.

Introducing v2 of our SFX model. Generate any sound effect directly from a prompt, via UI or API. • Higher SFX quality • New: seamlessly looping sound effects • Maximum duration extended from 22s to 30s • Sample rate increased from 44.1kHz to 48kHz • Refreshed SFX Library

Murió Edgardo Cozarinsky, uno de los grandes artistas argentinos. En un mundo de estridencias, lo más bello e interesante de la vida emergía en una charla con él, champancito de por medio. Salud, querido Edgardo. Tu obra me acompaña desde temprano y lo seguirá haciendo siempre💔

🤖 AP, Prisa, Le Monde y Financial Times acordaron con OpenAI. La inteligencia artificial se servirá del material que produzca esa agencia y medios a cambio de pagarle por su propiedad intelectual para producir conversación. ¿El periodismo se rinde ante la máquina o se impone?

We launched NotebookLM, the new AI-powered research/writing tool I’ve been building with @labsdotgoogle, just over a week ago. It’s a genuinely new software experience for learning, brainstorming, and more. Here's a🧵recapping the news since the launch. notebooklm.google

Summary of Gemini's 60-page technical report. 1. Written in Jax and trained using TPUs. The architecture, while not explained in details, seems similar to Flamigo's. 2. Gemini Pro's performance is similar to GPT-3.5 and Gemini Ultra is reported to be better than GPT-4. Nano-1 (1.8B params) and Nano-2 (3.25B params) are designed to run on-device. 3. 32K context length. 4. Very good at understanding vision and speech. 5. Coding ability: the big jump in HumanEval compared to GPT-4 (74.4% vs. 67%), if true, is awesome. However, the Natural2Code benchmark (no leakage on the Internet) shows a much smaller gap (74.9% vs. 73.9%). 6. On MMLU: using COT@32 (32 samples) to show that Gemini is better than GPT-4 seems forced. In 5-shot setting, GPT-4 is better (86.4% vs. 83.7%). 7. No information at all on the training data, other than they ensured "all data enrichment workers are paid at least a local living wage." Full report: storage.googleapis.com/deepm…