Framer AI AgentsDesign and publish professional sites with AI
Most AI writers just mimic your tone. Insight Engine also makes the thinking sharper. Give it a rough idea, a link, or a PDF and it researches real facts, drafts 3 angles, and scores every draft against 29 anti-slop patterns before it passes. Humanize mode fixes your ChatGPT drafts without losing the facts. Ideas page pulls fresh angles from HN, Product Hunt, and arXiv. No scheduling, no OAuth, just sharper posts, priced per credit, not subscription.
Congrats on the launch, Salta! I was going to drop some harsh critiques, but that voice onboarding totally caught me off guard. It was somehow able to pull out the exact idea in my head and narrow it down to my tone without me having to guess or over explain how I want to sound. Love that it actually sounds like me and not ChatGPT. Awesome stuff.
@emmanuel_eromosele Thank you so much, this genuinely made my day!🙏 The voice matching piece was actually one of the hardest parts to get right, so hearing that it captured your tone without you having to over-explain means we're on the right track. Really appreciate you giving it an honest shot and sharing this!
This is exactly my problem. I write educational posts, and they always end up sounding like everyone else's — even when the idea is mine. So I really hope this works. One question: when the AI pulls facts from arXiv or HN, does it show me the links? I want to check the sources before I post. Good luck! 🙌
@daulet_imanaliev Great question, sources are not excactly shown, i will add it! we know that trust matters a lot for educational content especially. Would love to hear how it goes once you try it on one of your posts, especially curious if the sourcing holds up for your niche!
How does the scoring against those 29 anti-slop patterns actually work in practice. Like is it a fixed rubric or does it learn from what gets flagged as slop.
@enes8lsg Great question! Not a fixed rubric, it's an LLM critic that judges each draft against those 29 patterns plus your actual voice profile, so it's judgment based rather than a hardcoded checklist. Still tightening a few edge cases(like numbered-list structures that look clean but are still templated), but that's the core idea
I tried it with one of my drafts. I liked seeing the different angles, but my original note had a few connected ideas, so I wasn't sure which angle to choose. In cases like that, do you recommend focusing on one idea at a time before generating a post?
@vaishnavi_makode Good question! Yeah, generally recommend picking one idea per post if you want the strongest angle options, when there are multiple connected ideas in one note, the angle generation has to guess which thread to prioritize, so results get a bit diluted. That said, we're looking at ways to let you split a multi idea note into a thread or a series instead of forcing you to choose upfront. Appreciate you testing it with a real draft!
@saltanat Thanks for clarifying!
The 'AI Critic scoring against 29 anti-slop patterns' is a clever inversion — most tools add slop, yours actively hunts it. My honest question: since the critic is itself an LLM, doesn't it risk sharing blind spots with the generator (both trained on similar data)? Curious whether you've seen cases where a draft passes the critic but a human still reads it as templated.
Been using this one firsthand, not a launch-day flyby. The thing that sold me: I fed it a messy idea and it held my actual voice instead of flattening it into the usual ChatGPT smoothness, and the post it wrote about breaking writing rhythm actually broke its own rhythm. Rare to see a tool practice what it preaches. If you post on LinkedIn or X and hate sounding like a bot, worth 2 minutes of your day. Nice work Saltanat 👏
Most AI writers will happily hand you a confident draft even when the
idea underneath is thin. One that actually argues with you about the
thesis first is rare, and honestly kind of humbling.