Hot take: @Replit has already won the AI coding agent game. The world just doesn't know it yet
Been using it to build a wrapper around the @LumaHQ API to create a membership portal for our community
Fully custom DB, User login, live data sync, restricted content, etc 🤯
@pirroh@raymmar@Replit@LumaHQ V2 agent went to creative and destroyed my app had to go back to v1. Plus it has no checkpoints to go back to before it starts and as it goes.
@raymmar@Replit@LumaHQ Try it again sometimes next week, and let me know how it goes?
We already made some changes to strike a balance between ambition and autonomy.
@etarud I'm fairly confident it was an issue with my usage. My prompt didn't ask for <thinking> tags, but my task was categorizing the intent of other prompts and if that prompt contained that instruction, then it leaked into my prompt. If I delimit with <prompt> tags then it's good.
Unless I have a big in my code it seems that Google Gemini Flash 2.0 is including a <thinking> block in its response by default. Anyone else seeing this?
@etarud Yeah, the more I debug it, it does seem inconsistent. My theory now is there is something in the context I'm putting in the prompt that is triggering it.
Does Gemini do context catching automatically? @OfficialLoganK I’m not a dev but building an app with replit agent and using Gemini flash 2 thinking as model integrated to the app and need to see if it’s automatic or not.
¡Buenas tardes!
#NovedadVial en el departamento de #Magdalena, se reporta cierre temporal en el sector La Caribeña. #EnDesarrollo
Trabajamos para mantenerte informado.
I've always dreamed of having my own watch collection. The Jacob & Co. Flight of CR7 & the Heart of CR7 takes inspiration from some of my most iconic moments on the field. I hope you like them as much as I do.
We're excited to introduce Keak, the first AI agent capable of continuously improving your website!
Our variations win over 80% of tests and getting your first test launched should take less than 5 minutes.
Try it for free: Keak.com
@etarud@kleffew94 It seems to be fixed now, just sending another message today it resumes. It's still not perfect, in one chat it went into an infinite loop of just asking I'm going to make x changes should I proceed over and over, but in another it actually started working again.
What do you do when replit agent gets stuck on a task, keeps spinning, stops responding?
They suggest creating a new project but I’m trying to make this project since days and it always gets stuck.
Excited for some football today! With my @Replit built app, I'm tracking the probabilities of NFL teams making it to each phase of the playoffs. Check it out!
nfl-team-playoff-tracker.replit.app
@kleffew94 Just straight up stops responding after I answer a question. The spinner gets stuck. If I kill and relaunch the app I can send another message but nothing happens and spinner stays stuck.
@pirroh@amasad@TimCodesStuff@Replit@zhenthebuilder Hi Michele, I have had my Replit agent stuck for several hours. I ask something, the message gets sent, and nothing happens. I send another one, and it's the same. What can I do to fix it?
BREAKING: The MacOS 2.0 ChatGPT app upgrade is NOT a "Nothingburger" as it seems at first glance. Check this new UI 👇
It already comes with a properly working conversational UI you all have seen on the OpenAI demo.
- It will very likely always appear in a separate window which you can drag around (unlike how it is working now).
- It will have access to camera and screen sharing options.
- Screen sharing will allow the ChatGPT app to see everything that is happening on your screen.
- A new voice mode will have a different icon (not new).
Saying this again and again - this app will be a HUGE power user tool and work aside by you as your "true" copilot.
Disclaimers:
- A new voice and vision model is not available yet
- This UI is hidden under feature flags and not available for the public use
OpenAI is making GPT-4o, GPTs, DALL-E 3, and data analysis tool free for everyone...so why keep paying for ChatGPT plus? 🤔
Two reasons:
- ChatGPT+ users get 5x the rate limit (messages you can send in certain time period.) so far our community is reporting this gives you about 80 messages per hour.
- The advanced capabilities they showed off in the Spring Update like the Live Vision feature are only going to be for ChatGPT Plus users.
So yes, OpenAI is making “GPT-4o level intelligence free for everyone” but there are a bunch of asterisks on that.
ChatGPT+ is still an absolute steal at $20/month for all the features and early access you get when compared to free users.