Alex Galinos’ Post

So as expected the OpenAI launch was a regular drag drop tool to build agents. Two things to take away from this. 1. The AGI dream is over. If you remember, the whole pitch was that one model will do everything. Now apparently you have to call different existing SaaS companies to get things done. 2. I know ChatGPT has distribution, but if the companies give up their customer base to ChatGPT, they will struggle later. It's better they include the chat window in their own apps. Any other way is not the right way. Finally, the less said about the drag drop system the better. Real AI systems don't need drag drop. You should be able to just specify.

Big News: OpenAI just dropped a browser called Atlas, and it’s not just another tab. What Atlas can do • Instantly summarise any page. • Remember what you read - yes, your browsing history becomes context. • Book tables, fill forms, shop, test code - all from the browser window. • Run cross-tab workflows like your personal digital concierge. Search isn’t just about keywords anymore, it’s about context and action. Chrome has 4 billion users. ChatGPT has 800 million. Both massive, but one’s built for doing, not just finding. (Source: public figures) The $250 billion search-ad model? It’s ripe for disruption Why it matters ? - If you’re creating AI tools, SaaS, bots this shifts the unit of value from “traffic” to “task-completion”. - If you’re building content/momentum your play isn’t just “get views” anymore, it’s “solve flows” inside experiences. - If you’re pursuing passive income — imagine owning pieces of actions (bookings, form-fills, workflows) not just clicks. TL;DR: Search as we know it is evolving fast. The winners won’t be the ones who rank highest, but the ones who get actions #AI #SaaS #DigitalHustle #BuilderMindset

What’s one way you’ve used AI to simplify a task? I’ll go first: At Thumbtack, our Support Advocates often got the same painful question from pros: 👉 “How many more reviews do I need to reach 5 stars? 4.9? 4.8?” Answering it was a slog: open the profile, click to view as customer, click to open reviews, manually count 1-star reviews, count 2-star reviews, repeat… then dump it all into a calculator. A 10+ minute process every time. I didn’t know BigQuery beyond copy/paste, but I partnered with ChatGPT. In about an hour of back-and-forth (“Is this the right file?” / “Send me a screenshot”), we built a lightweight query: Paste it in → add the user ID + target score → run → get the answer instantly. What took 10 minutes now took 10 seconds. It cut a major time sink, improved ACH times, and made life easier for the team. I went on to make several other helpful queries after that one little project unlocked an entirely new skill for me. You can read the full article about it here: https://lnkd.in/gxaY8dNn Your turn! #ArtificialIntelligence #AIatWork #ProcessImprovement #CustomerSupport #Operations #Automation #Innovation

Radix AI analyzed 10K+ SaaS prompts across ChatGPT, Perplexity, and Google AIO — and G2 continues to lead the way ➡️ G2 citations jumped to 9.31% (from 2–4%) 💡 1 in 5 product discovery queries reference G2 content 💡 “Best Of” guides = 56% of citations 💡 “At a glance” sections get cited more than any other page part More third-party validation on G2 is a must-have. Full article in comments 👇

Everyone’s talking about agents. But few are building the systems that make them useful. AgentKit is a new framework from OpenAI that lets you connect GPTs directly to real tools and APIs, so your AI can take action, not just chat. AgentKit just dropped...and I saw a gap: no easy way to build new workflows without diving into the logic manually. So… I built a GPT that does it for you. You tell it what you want to automate, and it maps the steps, defines the logic, and outputs a ready-to-run AgentKit workflow plan. Think of it like a workflow architect inside your ChatGPT. If you’re experimenting with AgentKit and want a head start, comment “AGENT” below and I’ll DM it to you. (make sure we are connected) #AgentKit #AIautomation #GPTBuilders #workflowautomation #buildinpublic

𝗠𝗶𝗰𝗿𝗼𝗕𝗿𝗶𝗲𝗳: Auto-Tag Google Drive Files by Topic 𝗤𝘂𝗶𝗰𝗸 𝗪𝗶𝗻: Find the right files instantly without manually naming or sorting them. 𝗛𝗼𝘄-𝗧𝗼 (𝗧𝗼𝗼𝗹 𝗦𝘁𝗮𝗰𝗸): Connect Google Drive to Zapier → trigger on “new file uploaded” → send file title and text to ChatGPT API → auto-generate and apply topic tags like “Finance,” “Legal,” or “Design.” 𝗜𝗺𝗽𝗮𝗰𝘁: Cuts search time by 70% and keeps shared drives organized automatically across teams. 𝗝𝗼𝗶𝗻 𝗡𝗼𝘄 👉 Subscribe for more AI workflows → https://lnkd.in/gUHhUhQ4 𝗗𝗶𝘃𝗲 𝗗𝗲𝗲𝗽𝗲𝗿: We show exactly how to train AI tagging prompts to match your company’s unique folder structure and naming rules. #MicroBriefs #TrivioAI #Productivity #Automation #AIWorkflow

OpenAI's leaked "Agent Builder" dropped today and most people are missing the real story. Everyone's focused on the no code drag and drop interface. That's not the play. The play is OpenAI commoditizing the entire workflow automation market. What they're actually doing: → Making Zapier's core value prop obsolete → Putting n8n on notice → Turning complex AI workflows into consumer products The leaked features everyone's hyping: - Visual workflow builder - Prebuilt templates - GPT 5 mini integration - No technical skills required Translation: They're going after the $10B workflow automation market by making it free (or close to it) for ChatGPT users. The question nobody's asking: If OpenAI can offer this as part of their existing subscription, how do standalone workflow tools justify their pricing? What matters today isn't whether Agent Builder is "smooth" or has a "polished UI." What matters is the pricing model and who gets access. If it's available to all ChatGPT Plus users, every workflow automation SaaS just became a feature, not a product. #AI #Automation #OpenAI

NEW: Plug Ahrefs straight into your AI Connect ChatGPT, Claude, or Copilot, and they'll pull fresh Ahrefs data and analyze it for you – powered by our new remote MCP server. Docs + setup guides:  https://lnkd.in/gAn89eSB Watch a quick demo 👇

💡 Your GraphQL schema is closer to being AI-ready than you think. With a few lines of code, your schema can become a set of AI-ready tools: typed, validated, and instantly usable by ChatGPT or Claude. No new UI, no marketplace friction. Just clean, schema-driven access for AI. We built it, tested it, and proved it works with 𝐒𝐚𝐥𝐞𝐨𝐫’𝐬 𝐆𝐫𝐚𝐩𝐡𝐐𝐋 𝐀𝐏𝐈 + 𝐅𝐚𝐬𝐭𝐌𝐂𝐏 𝐂𝐥𝐨𝐮𝐝. 🔥 If your data lives behind APIs, our guide shows how to bridge that gap – from GraphQL schema to AI-ready MCP server in minutes. 👉 Read the full article: https://lnkd.in/d9iH-jCY #GraphQL #Saleor #FastMCP #Schema #TechGuide #DevCommunity

We’ve just built the Text 𝐌𝐂𝐏 𝐬𝐞𝐫𝐯𝐞𝐫 --> a new way to connect AI assistants directly into Text. Why does this matter? Because now both our own 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗶𝗻𝘀𝗶𝗱𝗲 𝗧𝗲𝘅𝘁 and external assistants like ChatGPT, Claude, Cursor, VS Code, Windsurf… can securely work with tickets and chats: no hacks, no extra middleware. Here’s what it unlocks: ➡ search and filter tickets or archived chats without leaving your assistant ➡ pull full transcripts into Text Copilot (or your other favourite AI assistant for analysis) ➡ ask natural questions like “Show me all billing issues from this week” and get structured results back ➡ keep everything secure --> your permissions in Text are mirrored exactly, so there’s no overreach In short: MCP makes AI assistants actually useful inside your support workflows instead of just sitting next to them. I’ve already used it to summarize billing conversations and surface common pain points in seconds - tasks that used to mean exports and manual digging. Curious: if your AI assistant (or Copilot in Text) could safely see your tickets and chats, what would you use it for first? 👉 More details and setup guide here: https://lnkd.in/e9ac3DRz