Loading...
Loading...
I have been a die-hard fan of the #openai models for coding when I code I #Vibecode but I use an IDE like VScode. Yesterday @google announced the release of #Gemini Pro 3 and also an idea that they are calling #antigravity , I had a project that I needed to get going last night SO, I thought after hearing that Gemini Pro 3 was rather good, I thought I would give it a go. I have to say it is seriously good I am impressed with how they have built and implemented and their agent but also the quality of the output from Pro 3. While there are some pretty hardcore usage limits with 4-Hour cooldowns which for me is so very painful (reminds me of GPT models from 2 years ago). it is actually something I would consider paying for, purely because of how good it is. So if you are looking to Vibe code your own apps then it's absolutely worth having a look.
Performance Category
Above Average
Score
4.7/5
Shares: 5/5
Comments: 5/5
Retention: 5/5
Views: 5/5
Likes: 5/5
Followers: 5/5
Script: 3.0/5
Total Views
23437
Likes
842
Shares
80
Comments
70
Duration
2m 53s
For You
19,570
83.5% of views
Search
2,414
10.3% of views
Personal Profile
984
4.2% of views
Others
305
1.3% of views
Follow
164
0.7% of views
Sound
0
0.0% of views
Views
Likes
Shares
Comments
For You Traffic
Profile Traffic
Search Traffic
Non-Followers
80.0%
18,750 views
Followers
20.0%
4,687 views
16.6% of followers reached
New Followers
170
Performance vs Median
No transcript available.
No scorecard available.
Run the script scorecard evaluation to generate insights.
Run the AI evaluation to identify actions and analyze their impact.
No tips available.
Run the script scorecard evaluation to generate actionable tips.
Barber87Qa roles absolete the ?
Mzticccan I use my github copilot with it ?
no need.
Imran RykliefBro dark mode exists
ewwwww... I can't stand dark mode.
n8danielsBeen using playwright to test my front end but what you are explaining seems very impressive!
Awesome, glad it's helpful
AlgoBrushesArtFor me the only thing that saves it is the autonomous browser thing, Gemini really hates following instructions, I think the future are tools like GitHub spec kit, you can use different models for different phases and use almost any with the task phase, I usually ask different models to review the specs and get more insights. What I like the most about spec kit is being able to use with Antigravity, Claude Code, Gemini CLI, OpenCode and Cursor at the same time
Yeah, the limitations of the timeouts just got to me too much. So I finally migrated over to cursor where I just switched between GPT 5.2, 5.1, and Gemini all the time. The composer one in Cursor is actually pretty good as well. So I switch between all of them.
WoodyStillLaughsbeen playing with kiro for a few days and it's fah king dope!!!
Nice
sky_sailRate limits are horrendous, cooks out in 10 minutes
I know, so freaking painful.
Morten of the North 🇨🇦🇳🇴It's a fork of Windsurf.
I assumed Windsurf was a VScode fork, but I'm wrong on that! After reading their reasoning behind the built, makes total sense why they fired WS and not VS. Also, Hi @mor10web long time follower.
einteresting
I thought so too
Abbot SkinnyCaveat, it tends to loop out endlessly if the implementation plan steps are contradictory- meaning i watched it frustrate itself trying to not to do something stated earlier but mapped later. Had to stop its thinking and say “Bro, just remove that part of the plan you’re straight trippin” And G was like ok let me build a - and i was all “No just stop, you’re all over the place” Its got a bias to waste token essentially.
Brennan McDonaldI am looking forward to be able to test Google Antigravity properly without hitting Gemini Pro 3 (high) limits - it seems like a codex/claudecode peer
It's so painful!! I'm getting maybe 45min or runtime every 12 hours :(
Growgamilove your vids, very insightful!
Thanks!
Chang NoiYou're missing something. I just ran this for the last two days and I'm not usually using Google's atmosphere. I've played with it but mainly use VS code I use a little bit of cursor. Use some Claude code, but I went through a test run yesterday and I did two things one. I just created a test app very structured, but ran a test app. I first ran it through AI Studio then I hit up fire code to implement a few things. And that was fine so then I went down to the local level downloaded gravity and I think it's great. I still feel comfortable in a VS code set up right now, but for Google, I would run out of the antigravity and there's a interesting little thing that happened. Maybe others will investigate this. I told Gemini to help me create a quick extension just because I have issues seeing colors and I don't like what the other extensions do and I don't wanna spend time trying to mess with those. So I told him I'm gonna use gravity and I wanted to help me create a basic extension what is the PRD what is the workflow and how can I implement this. And it just cooked .. but I let it know that I was using zero gravity and it did something pretty crazy. I also told her that I was sick of it calling everything Gemini so I just sarcastically told it I'm calling you guys, Gemini, Gerry, Gemini and Gerald. Now I know that GS code has agents has the chat, etc.… But it's always been kind of difficult for me to realize which AI is in acting where etc. but this is what Jim Gemini three made for me and now I'm wondering if this is built into amplify through the Google domain.? I don't know but take a look at this READme file. I think this might be really interesting in implementing some extension actions that can connect within your Devon environment. I don't know maybe this is already existed forever with people who are familiar with coding and Dev with Google, but I found it pretty interesting.
Chang NoiThe Gemini three in the task Iran over the last 48 hours has been absolutely sufficient, completely capable of managing steps task protocols and expectations specified by me and if you use the learning mode with Gemini three while you're working through a coding experience or a lab or some sort of reference manual, it's amazing even when you start to ask questions about other things related to it it'll draw you back into the objectives and say hey let's get back on track with this
Chang NoiOn the basis of rate limits set up a Google Cloud account and set up a Google Dave account and you get $300 free of credit and I'll let you use the Gemini three and you're not getting throttled. I wasn't throttled yesterday and I think I worked two straight days without stopping.
Chang NoiOh one last thought also go to GitHub and grab the Gemini API cookbook. Then you'll start really getting the power of this 3.0 API.
JasonI was lucky enough to time my one year Cursor subscription before they removed the unlimited plan. So I've been hoping local models catch up in the next year to do this all self-hosted. But Antigravity fixes a lot of Cursor's issues, so I'm curious what their paid plan ends up being like.
On nice catch on the unlimited! Yeah the limitations on the model are really hard to make work, I'm only getting 1.5 hours of work in a a 13 hour day... So almost unusable, and trying to switch between this and my usual gpt codex in vs code.. Going to cause issues
JasonIf you haven't used it recently, plan mode helps a lot in the newer cursor. Also I've found a memory bank prompt in rules works better than mcp memory servers to keep the project on track. Opening a fresh chat for each thing I do, knowing it will pull context from my memory back.
I just hate to pay subscriptions to just be able to use my own keys and not their markup fees on tokens.. ill take another look though.
MindlessPuppetzI’ve been using Cursor for a couple months. Loving it. Very similar to VS Code but AI integrated into it. Can use different agents.
I tried cursor, I just don't like the subscription
MindlessPuppetzI used it for free for a bit but the $20/mo is honestly worth it for me. It has been awesome using it for LUA/web development
I'm always going to try avoid another subscription if I can.
Alex Ecomglm 4.6 best option
MiraDiary.this vid explains ai agents well. workbeaver works similarly, i just show the workflow and it performs it afterward, makes automation feel natural.
Thanks
rose(✿^‿^)🏳️⚧️🏳️🌈does it give me extra stuff for having Gemini AI pro subscription?
Nope, there is no upgrades, I would have paid already :(
rose(✿^‿^)🏳️⚧️🏳️🌈ugh, I hate that, honestly it's so annoying how often I get very similar limits or access as the free tier, but if you want proper access to alot of their stuff either you setup API payments or you pay $300 a month for ultra it's really annoying to me personally
Totally agree. I'm assuming that they will change it soon enough.
Guerrilla GorillaHoly moly
I know!
Evan Jemergent has been doing this already.
I'll have to take a look
Raven Inkthere's no reason to fanboying to only 1 company's model, the AI race still going hot
Oh yeah, totally, not fanboying at all, just impressed with their jump ahead. I'm still going back to 5.1 codex in vs code all the time.
BlahThe rate limits make it unusable at the moment
It's super painful yeah.
I should have been far more sarcastic when I quoted their "Generous rate limits"
BlahIt is indeed. But it’s a cool tool
RobinBanksGoogle is the leader in Ai 👍 openai will be lucky to make it 2 years
I don't know about openai, but yeah Google took its time but this really hits well.
Mxrktesting now
How did you go?
MxrkI ran into a problem with the AI assistant extension not working in the editor view because most of my VPS is are on shared hosting servers that don't allow you to change your CPU type and found a workaround where you can use the AI agent system window to actually still interact with those certain VPS's.
ishtiaqueahmed796by the way I have been using anti-gravity since yesterday. it's very impressive.
Apart from the errors, but I'm assuming high usage is the case.
ishtiaqueahmed796yeah I have noticed that retrying works. will be interesting to see what you are doing with it. happy to connect offline.
ishtiaqueahmed796browser tool is also available in roo code and through playwright MCP.
Mxrkis roocode free or something? or more privacy focused?
ishtiaqueahmed796roo code is free in vscode. you have to pay for the models though.
-ANTHONY-built for agentic devs who care about systems design but just cant be bothered to focus on syntax.
For building applications to run in your homelab, it's amazing.
JoshI don’t want to know that!!!! I want to continue to avoid google 😭
I know!!!!! I was so pained that it was actually really good.
simbalazy bum u couldn't do this ur self
Which bit?
UMNZ 🇳🇿Google know exactly what they’re doing and this is not for “vs code” peeps. This is going down a different path. AWS and Google are going a different direction to the rest.
yeah, Its 100% going for agentic developers with human input/guidance. Im here for it, but the rate limits are PAIN.
DaoentiI mean, it’s similar to VSCode because it’s absolutely a codium based IDE, just with their custom plugins in it. It’s the same thing Amazon did with Kiro.
Yeah, they are all just forks of VS code, or a plugin, why reinvent the wheel I guess.
DaoentiYup, I may try this one out at home, I use Kiro at work and have really liked it but I’m not a vscode person and it’s been rough switching to a codium based IDE. It sounds like the implementation plan in Antigravity sounds like what I love in Kiro, targeted steering docs and spec files and it’s just something I can’t live without anymore when using AI to write code. When they released Kiro CLI two days ago I was ecstatic because I can finally go back to nvim.
Nice, I feel too hands off with cli only, I guess I'm just more visual.
Total viewers and likes aligned with spoken words.