Oleg Zankov

🚨BREAKING NEWS: Replit just dropped Agent 3. And builders are about to lose their minds. It’s the new operating system for builders. Agent 3 changes the rules. Here’s how smart builders will use it: 1 – Build end-to-end inside Replit ☑ Prototype apps in minutes ☑ Test & fix code with an actual browser ☑ Deploy without leaving the workspace No context switching. No tool hopping. 2 – Automate what drains your time ☑ Slackbots that pull data on demand ☑ Telegram reminders for habits & tasks ☑ Daily email reports generated + sent automatically You describe the flow. Agent 3 builds and runs it. 3 – Unlock hours of autonomy ☑ Runs for 200 minutes on its own ☑ Checks its work, fixes bugs ☑ 10x more reliable than before It doesn’t just generate. It collaborates. 4 – Scale without 3rd-party crutches Competitors need add-ons for auth, DB, or deployment. Replit ships them built-in: ☑ Auth ☑ Database ☑ Deployment ☑ Monitoring All under one roof. 5 – Protect your focus Every builder knows the grind: Prototype → Test → Debug → Deploy → Repeat. Agent 3 takes that loop off your plate. You focus on ideas. It handles execution. Your workflow becomes: Idea → Describe → Ship. That’s autonomy. And for those ready to try? 👉 https://replit.com/agent3 Autonomy for All. Repost ♻️ so more builders discover this.

870

139 Comments

🚨 Tested 100+ AI Productivity tools so you don’t burn months figuring it out 👇 Last 30 days, I went deep into every legit AI tool teams are actually running in production. The question every client asked: which tool should we bet on in 2025? Here’s the signal (not the noise): 1️⃣ Chatbots — It’s no longer just GPT. DeepSeek 🛑 has the dev crowd. Claude 🛑 rules long-form. Perplexity 🛑 quietly killed Google Search for researchers. 2️⃣ Coding Assistants — This category exploded. Cursor is eating share fast. GitHub Copilot is now table stakes. Niche players like Qodo and Tabnine finding loyal users. 3️⃣ Meeting Notes — The silent productivity win. Otter, Fireflies, Fathom save 5+ hours/week per person. Nobody brags about it — but everyone uses them. 4️⃣ Workflow Automation — The surprise ROI machine. Zapier just embedded AI. N8n went AI-native. Make is wiring everything. This is the real multiplier. Biggest gap? Knowledge Management. Only Notion, Mem, Tettra in the race. Feels like India’s UPI moment waiting to happen here. Unpopular opinion: You don’t need 100 tools. The best teams run 5–7 max — per core workflow — and win on adoption, not options. What’s your take? #AI #SaaS #Productivity #Startups

821

16 Comments

There's a Shopify performance hack that made an Indian devshop pretty famous on UpWork a few years back. Google Page speed showed 99% score. Lightning fast - sites were loading in milliseconds. They could deliver this result in under a week, and only charged $200. What a steal. How did they do it? They removed all the content from the website *BA DUM TSSS*. Told you it was a steal. They made a killing before being discovered. They put a special script on Shopify stores that would detect if a Google bot was scanning their site. If so, it would render an empty page. We're talking header, basic menu, and then maybe the footer. Since 80% of the content was gone, it's a 5x speed increase. Wow, so gr8. Much fast. Do you know what else relies on Google bots scanning your site? SEO 🤦‍♂️ Not only were they lying to customers saying their site was faster. Which was wrong, since there is no tangible difference to real users. They obliterated any Google ranking and search-ability. There are 2 really important lessons here: - site performance is mostly about removal (ecom is heavy) - you can tailor your site to different perspectives (focusing on this 👇) - (bonus 3rd, Shopify scores > 70% should be scrutinised, very rare) When you first load a website, the user can only see the length of their screen. On mobile, that's really not a lot. Probably just your PDP carousel and title. But most of your page load is being spent on off-screen: - review carousels images - cart upsell apps - listicle like information underneath product This is a bad experience for users. You may have some amazing videos or content loading off-screen, but they don't know that. To them, it's just a slow, clunky vitamin site. It obliterates cold traffic. This is where we use Lazy Loading. Lazy loading it's a way of making the website feel faster than it really is - by being prioritising load. In some occasions, it can reduce network congestion entirely by spacing things that load. How to implement lazy loading: 1) Load your most popular PDP page, at the most popular screen width (if you don't have dimensions, just use mobile on a iPhone 12), on your most popular browser. 2) Go to Google Page Speed insights and run a test on that page 3) Pay attention to EVERY image, script, css file, or font it flags as "large file", or "render blocking". If you can't see that on the screen, and it's not doing anything important in the background - lazy load it! Images can be lazy loaded (which means they wait until you scroll to them before loading), and scripts can be defered (which means they wait until after the page has rendered to fire). Both of these can be done with HTML, and are easy for your devs to implement. Just give them a list and they'll handle the rest. 5 mins of admin, seconds shaved off page load. It's a good trade

50

18 Comments

(Whispers: Schema is still valuable, ya'll. For SEO, and LLMs, too. It's not the magic bullet some GEO "experts" are claiming it is, but it can still help. Assuming it's inline in your Response HTML, anyway - but that's the setup for many, many instances. And when that's the case: schema is... text. Nice, structured, summarized, and (dare I say?) "chunked" text. Available to all all those bots working to decipher your page contents...)

62

31 Comments

UPDATE: Assist Rate Metrics You can now track how well your AI Agents assist with the issues they're assigned to. An agent's assist rate is the percentage of assigned issues where it either... → runs a runbook → responds to the customer ...and you can see an exact breakdown of how many times it did each (or both). This helps evaluate the agent's performance more holistically and identify where to tweak its configuration. Enjoy!

31

6 Comments

There’s a seismic shift happening in how to distribute on the web ... from organic search (SEO) to generative engine optimization (GEO) & answer engine optimization (AEO). At QuillBot, we’ve spent years building the writing tools people rely on and making sure they can find them. That work now drives over 30 million clicks per month and a billion impressions per month. Now, with generative & answer engine optimization (GEO & AEO), we’re also ensuring that when people use ChatGPT, Gemini, Claude, QuillBot Chat, Perplexity, and other AI tools for help with multimodal communication, they find helpful information about our products in the answers. Our historical focus on providing solutions through traditional search channels still matters, but if we want to provide our tools to everyone, we need to get cited, quoted, and referenced by AI, not just clicked in search results. Some practical moves (hat tip to Ethan Smith via Reforge): - Target long-tail questions and answer them in-depth. - Fully explain all of your product’s features and integrations. - Create YouTube videos featuring your product’s use cases. (Perplexity, in particular, loves to cite YouTube.) - Engage on Reddit, and be sure to add value, not just self-promotion. We’re applying these strategies at QuillBot to provide as much helpful content to answer engines as possible and seeing promising results. We’ve shown up in 141,000+ AI Overviews so far—and that number’s growing every month. I want to learn for myself as much as I can and share my learnings with those most interested in shaping the future of search. (1) Drop a comment and tag someone you think I can learn from. (2) Follow me if you want to learn with me about distributing a product people love via SEO, GEO, and AEO in an AI-forward world.

149

42 Comments