AI Agents will dramatically expand the size of the software market. Here's how that will work.  Traditionally, software companies are stuck within the constraints of existing IT budgets, with IT expenses running in most companies somewhere between 3-7% of revenue (with tech and banking often a bit higher). This has always introduced a natural ceiling on the amount of spend for most categories of software. But in an era where the software, because of AI, is *solving* the problem for the customer and not just *enabling* a solution to the problem, the ceiling gets blown up. In an AI-first enterprise, AI Agents will: help marketing teams will spin up campaigns faster in all regions; code and test software for engineering; answer and triage first layer of support tickets; scale outbound campaigns and generate leads; automatically review and work through contracts; and so on. None of these outcomes traditionally came from IT spend. This then directly leads to software TAM growth. Just take a micro example of legal use-cases as an easy case in point. In the US, the contract management and ediscovery software categories are a few billion dollars each, give or take. However, the size of the legal services market in the US is somewhere around $400B, nearly 100X larger than the related software categories. If AI made the legal services operations even 20% more efficient (which is likely an understatement in the medium run), the software spend in this space could very easily grow by 5-10X. You can apply this logic to basically any category of work, and the math is similar. Importantly, the spend is not inherently going to be zero sum with what we spent on before. Net new dollars for AI (not replacing labor) will appear in many areas: startups and small businesses will go after problems they couldn't afford before; teams in large companies can scale out an operation far more than they would've otherwise; and teams that maybe had a business requirement but not enough budget or weren't prioritized before, can now solve a problem more quickly.  Going forward, a company will simply decide how much productivity they want to spin up in the form of AI Agents, and they can just modulate quickly based on the ROI of whatever Agents they're using. Because of this flexibility for scaling out work, the new-use cases it now solves, and the ability to go past the typically limited IT budgets. AI Agents will make software markets much, much larger.

What's the best drag-and-drop way to build AI agents right now? - Langflow - Flowise - Gumloop - n8n or something else?

Build Your Own Deep Research Agent with AutoGen @OpenAI just released its deep research agent - But did you know you can implement similar functionality with a few lines of code with @pyautogen ? Similar to the operator agent, here is an implementation showing similar results on the same tasks with the deep research agent - all in less than 90 lines of code (mostly agent prompts, excluding tools). Implementation TLDR (quite simple) - 3 Agents. AssistantAgent that performs web searches (using a websearch tool), a Verifier agent that ensures research quality and completeness, a Summary agent that provides detailed markdown summaries. Each agent uses gpt-4o. - A GroupChat pattern (an LLM is used to determine the order in which agents act/speak) Notes: This implementation explores an autonomous group chat pattern (there are some contro/reliability tradeoffs here), and the prompting for agents could be improved to finetune behaviour. #multiagentsystems #multiagentbook References [1] AutoGen Core v0.4 Documentation [2] AutoGen AgentChat v0.4 Documentation [3] A beginner friendly introduction to multi-agent systems -

Agentic report generation is going to be a headline use case among every enterprise this year, and Deep Research by @OpenAI proves it. General tools like deep research capabilities will be enough to satisfy general information requests. But true adoption in the enterprise will require the following capabilities which still need development: 1. Proper templates for use cases (e.g. filling out a questionnaire vs. generating a financial report). Ideally can generate in direct file formats like PDF, powerpoint. 2. Indexing a proper corpus of offline data - giving the agent an effective “unlimited context window” over your knowledge base, whether with RAG techniques or other tool calls. 3. Rich human-in-the-loop editing and validation in a way that can be domain-specific. Generation in a legal setting might imply different UXs than generation in an eng/product setting. Would require rich integration with different tools. We’re working on these capabilities with LlamaReport at LlamaIndex - with API accessibility to enable developers to build deeply custom, domain-specific report gen workflows. Come check it out! Blog: Come chat:

Introducing deep-research - my own open source implementation of OpenAI's new Deep Research agent. Get the same capability without paying $200. You can even tweak the behavior of the agent with adjustable breadth and depth. Run it for 5 min or 5 hours, it'll auto adjust.

Sam Altman about upcoming agent (swarms) in different fields.

Excited to introduce Icon, The First AI Admaker. We’re backed by Peter Thiel’s Founders Fund & execs of frontier AI labs like OpenAI, Pika, & Cognition. Icon ( is like ChatGPT + CapCut, but for making winning ads with AI in minutes. How it works: 1. Icon looks at your video library & tags scenes (e.g. "close-up," "unboxing"). These scenes become reusable clips used as lego blocks for making ads. 2. Prompt Icon’s AdGPT to generate scripts focused on specific angles & audiences. 3. Icon finds perfectly matching clips for every script scene & generates an ad that is 80-99% complete. 4. Make edits with our CapCut-like video editor until you're happy. 3-person creative teams make 30 ads per month. With Icon, they make 300. We co-built Icon with $100M+ revenue brands like Ridge, Jones Road, Immi, Backbone, & MUD\WTR to solve big pain-points: 1. Making lots of ads is extremely painful. Icon helps you automate the tedious parts of scriptwriting, scene matching, video editing, audience research, UGC creation, & more. 2. AI-generated ads look like trash. Icon remixes your existing footage into new ads, matching your production quality & maintaining brand aesthetics. 3. Existing solutions charge $2K-$30K/month for: 🔍 Competitor ad spying & cloning 📈 Creative analytics 📹 Custom & stock avatars for AI UGC 👥 Audience research 📁 Video storage & tagging Icon does scriptwriting & video editing on top of everything above for just $999/year (their margin is our opportunity 😉). If you’ve made it this far, we have a surprise for you 👇🏻

Build custom LLM apps with a drag-and-drop interface

MindSearch: LLM-based multi-agent web search framework

The biggest lie in AI: "Just add an agent and it'll work." The truth: Agents need: • Clear task definition • Role-based constraints • Proper orchestration That's how you build real solutions.

this is f****** insane from prompt to functional product integrated with @stripe for payments in less than 4 minutes, without coding. @donvito @SoftgenAI

Big Legal News for LLM AI Agents: Users Have the Right to Reverse Transactions if the Agent Did Not Provide a Way to Prevent or Correct Errors! (well... it was big news in 1999 when the law was passed, but its news to everybody today!). Read all about it: