Blog
Notes on the new Deepseek r1
Before anything, let’s bow to Richard Sutton; he was so early for this. Pure RL, neither Monte-Carlo tree search (MCTS) nor Process
Notes on the new Deepseek v3
Deepseek released their flagship model, v3, a 607B mixture-of-experts model with 37B active parameters. Currently, it is the best open-source model, beating
How 11x added $4.2M in enterprise deals and saved 380 engineering hours with Composio
How 11x Added $4.2M in Enterprise Deals & Saved 380 Engineering Hours with Composio $4.2 M Increase in ARR Pipeline 34%↑ Improvement
How Composio Helped Gorilla from UC Berkeley Ship AgentArena
How Composio Helped Gorilla, UC Berkeley Ship AgentArena saving 100s of Engineering hours 80% ↑ Increase In Tool Calling Accuracy 100+ Development
Gemini 2.0 vs Flash vs OpenAI o1 and Claude 3.5 Sonnet
Google has finally woken up and decided to drop the bombshell Gemini 2.0, completing the AI trifecta. Google has launched two new
OpenAI o1 vs Claude 3.5 Sonnet: Which One’s Really Worth Your $20?
It’s been a week since OpenAI o1 was out of preview, and along with that, OpenAI has also introduced a new tier,
How Letta is advancing AI with long-term memory | Insights from Charles Packer
In this episode of the AI Agent Podcast, we were thrilled to host Charles Packer, co-founder of Letta and the original author
How Assista AI reduced their go-to-market time by 90% using Composio
Assista Accelerated Time to Market by 8x and Saved 100s of Hours With Composio 8x Faster Time to Market 230+ Engineering Hours
How Fabrile integrated Gmail and Google Drive in just 30 minutes, fueling their success at TechCrunch Disrupt
Fabrile by Ingram Technologies integrated Google apps in 30 minutes to showcase MVP at TechCrunch Disrupt. < 30 Min To Integrate Gdrive,
Notes on Chatgpt Search: Better than Perplexity?
After much speculation, OpenAI launched the Chatgpt Search feature. It can now search the Internet based on your query and find relevant
Tool design is all you need for SOTA SWE agents
Introduction Building reliable AI agents is hard, but it does not have to be. One of the critical concerns for large-scale adoption
Notes on Anthropic’s Computer Use Ability
Anthropic has updated its Haiku and Sonnet lineup. Now, we have Haiku 3.5—a smaller model that outperforms Opus 3, the former state-of-the-art—and
