Supermaven
@SupermavenAI
Followers
6,754
Following
21
Media
28
Statuses
295
Explore trending content on Musk Viewer
Vance
• 989520 Tweets
Alberto
• 405045 Tweets
Iraq
• 396956 Tweets
#เทควันโด
• 338682 Tweets
BEST WISHES FOR SAROCHA
• 186444 Tweets
Kursk
• 185447 Tweets
Wisconsin
• 155151 Tweets
AIPAC
• 139003 Tweets
GIVE VINESH SILVER
• 130391 Tweets
Duda
• 85454 Tweets
Japinha
• 68853 Tweets
Eau Claire
• 63946 Tweets
#ROBLOX
• 55053 Tweets
Augusto Akio
• 52674 Tweets
#erişimengelinikaldır
• 37170 Tweets
Leverkusen
• 34591 Tweets
Walthamstow
• 34175 Tweets
Pedro Barros
• 27887 Tweets
Home Affairs
• 22561 Tweets
Ana Patrícia
• 19663 Tweets
Quincy Hall
• 18270 Tweets
Havertz
• 16790 Tweets
Xhaka
• 14781 Tweets
INDEC
• 13889 Tweets
#الاتحاد_انتر_ميلان
• 12476 Tweets
Bon Iver
• 11340 Tweets
Pinned Tweet
We've trained Babble, a new model with a 1 million token context window.
Babble is 2.5x larger than the previous Supermaven model and upgrades our context length from 300,000 to 1 million tokens.
We are deploying the new model to all Supermaven users today.
26
29
517
Introducing Supermaven Chat: the best way for developers to use GPT-4o, Claude 3.5 Sonnet, and other chat models.
If you've ever copy-pasted code into ChatGPT, then Supermaven Chat is for you. Save time by using Chat to upload your code while staying in your editor.
25
17
225
Our Neovim plugin is ready! Check our profile for the link to try it
15
10
105
VS Code extension version 1.0.5
- Adds "Generate Commit Message" button in Source Control tab (Pro users only)
- Chat window no longer needs to reload whenever it's closed and reopened
- Added checkbox to control whether "Fix with Supermaven" is shown in Quick Fixes
15
3
97
We have a Neovim plugin ready (in beta). DM us or check our Discord to try it!
11
4
81
We've added a free tier to Supermaven!
It doesn't have our unique 300,000 token context window, but it still has a great model with unmatched latency. And it can see your current diff, so it understands what you're working on.
5
7
78
The Supermaven Jetbrains plugin has been submitted to Jetbrains for review and should be ready in a couple days.
We'd be grateful to anyone who wants to beta test the plugin - DM us if interested.
10
4
77
The Supermaven Neovim plugin has reached 256 stars! 65,536 soon
0
0
73
We deployed a new scheduler that reduces latency ~3x for the most challenging subset of queries:
4
2
70
We've advertised our 300,000-token context window. But can Supermaven actually use all those tokens to improve its suggestions?
In our new post, we evaluate Supermaven's long-context performance.
8
5
69
Introducing Supermaven:
8
11
58
Our Neovim plugin is ready!
We'd like to thank the community for contributing PRs to make the plugin more robust and easy to use.
5
9
60
Supermaven Chat now supports GPT-4o mini!
Pictured: GPT-4o (left), GPT-4o mini (right)
7
2
60
Supermaven 0.1.23 (VSC) makes a significant change to suggestions in the middle of the line. They will now display off to the right so your text doesn't move around as you're typing:
5
3
52
We're organizing a meetup in SF next Thursday July 18 at 5:30pm!
Come to meet our team, talk about the future of AI for coding, and give us feedback on Supermaven.
Link below-
4
4
51
Compared to our previous model, Babble has been trained on more data for longer, and is 2.5x larger. This makes the completions much better.
Despite its larger size, Babble is still faster than our previous model thanks to improvements in our serving infrastructure.
1
1
47
We benchmarked Babble to test its ability to use its 1 million context window. Running a "needle in a haystack" test, we find that the model can retrieve a "needle" string from anywhere in the context window, regardless of where it is hidden.
1
1
42
There is no longer a delay for indexing when using Supermaven with a new repository. Supermaven will now return completions immediately and run indexing in the background.
5
2
40
The Supermaven Jetbrains plugin is ready (in beta). We tried to DM everyone who messaged us about it but we got rate limited by Twitter, so you can access it here:
3
3
41
Babble has been deployed to all Supermaven users, and Supermaven Pro users will be able to use its 1 million token context window moving forward.
We hope you enjoy using Supermaven 1.0!
2
0
39
Besides Babble, we've made several improvements to Supermaven in the past month, including team billing, Supermaven Chat, and infrastructure improvements to ensure low latency.
We're calling this release "Supermaven 1.0" to reflect all these improvements.
2
0
36
Supermaven for VS Code 0.1.31 supports enabling/disabling Supermaven for specific languages. For example, you can disable Supermaven in Markdown files.
8
1
36
We're excited to announce support for Jetbrains IDEs!
Thanks to all our beta testers who have helped work out issues with the plugin over the past weeks.
4
3
34
In addition to .gitignore, Supermaven now supports .supermavenignore to give you additional control over which files are indexed by Supermaven.
Syntax is the same as .gitignore.
1
2
30
Supermaven Chat now supports 8,192 tokens for Claude 3.5 Sonnet
Good news for
@AnthropicAI
devs:
We've doubled the max output token limit for Claude 3.5 Sonnet from 4096 to 8192 in the Anthropic API.
Just add the header "anthropic-beta": "max-tokens-3-5-sonnet-2024-07-15" to your API calls.
155
262
3K
3
1
29
Comparing to our previous model at predicting randomly shuffled Python files from the PyTorch repository, we find that Babble outperforms our previous model at all context lengths and also scales better with increased context.
1
0
25
We made changes to the backend that speed up building prompts, giving latency improvements up to 30ms depending on your repo.
1
1
23
Supermaven Chat has hotkeys for starting conversations, uploading files, requesting edits, and automatically applying them.
Here's an example of using Supermaven Chat to add documentation to a function. We request the edit using Cmd+I and apply it with Esc+A.
3
0
24
Starting from version 0.1.33, Supermaven is disabled by default in files ignored by .gitignore, and these files will not be sent to our servers.
You can re-enable Supermaven in these files by clicking the status icon:
2
1
23
We added a Code Policy to our website to give customers additional clarity on the protection of their proprietary data. We do not train models on customer code.
1
2
22
Using Supermaven to write a simple toy program. You can see we get streaming completions from the Jetbrains plugin:
8
1
21
With Supermaven Chat there are no rate limits - if you exceed your free credits, you'll be asked to authorize an overage charge and if you accept you can keep using Claude/GPT-4 as much as you like.
6
2
21
Some users on Windows reported that the Supermaven extension would keep a blank terminal window open. We reproduced the issue and it is fixed in extension version 0.1.13.
2
1
19
A long context window doesn't just help Supermaven look up function definitions - it also lets us learn your personal style and write code the way you would write it.
@SupermavenAI
is the only AI tool I've paid for. I never got much value out of GH Copilot. It's just too slow. I am able to type out what's in my head faster than what Copilot could suggest. This is not so with Supermaven. Not only is Supermaven faster, but it writes like me.
1
0
5
2
0
19
Supermaven Chat complements our existing inline completion functionality - the fastest of any copilot - and is available in our VS Code plugin starting with version 0.2.10. Chat is coming to Jetbrains IDEs soon!
3
1
19
Here, we upload our file and request a change to it. Supermaven Chat automatically associates the model's response with the section of our file that it's editing, allowing us to display the diff and apply the changes in one click.
1
0
18
Do you use Vim or Neovim?
Vim
26
Neovim
111
I use neither
92
9
1
16
We made changes to Supermaven's handling of large diffs that should significantly improve quality and latency when making many changes or adding large files.
0
0
16
Supermaven Pro users can use Chat at no additional charge. There are no usage limits - if you exceed your monthly free credits, we'll charge you for the additional usage.
If you're a Free Tier user, you can use Chat by providing your own OpenAI or Anthropic API key.
4
0
14
Supermaven had an issue that would cause completions not to show in large files (200KB+). This is fixed in VS Code extension version 0.1.22.
4
0
14
We fixed an issue with our inference kernels that caused corruption of model state under load, leading to incorrect/degenerate completions.
3
1
14
VSC extension version 0.1.47 improves trimming of redundant lines in situations with trailing commas
1
0
13
There was an issue with incorrect handling of carriage returns that caused Supermaven to fail to return completions on Windows in some cases.
This issue is fixed. It is a server-side change, so no update is required.
6
0
12
We changed Supermaven to be less conservative in suppressing suggestions in the middle of a line. To get these changes, update to extension version 0.1.17.
2
1
13
VSC version 0.1.49 and Jetbrains version 1.24 fix an issue where Supermaven wouldn't work in repositories with no commits.
2
1
13
You could be the 3rd view on our Youtube channel
(Uploaded because twitter was compressing our video too much)
1
0
12
Our SF meetup is in 2 days (this Thursday)! We'll provide food and drinks.
We're organizing a meetup in SF next Thursday July 18 at 5:30pm!
Come to meet our team, talk about the future of AI for coding, and give us feedback on Supermaven.
Link below-
4
4
51
0
2
11
There was an issue where Supermaven would consume a large amount of system memory when the repository contained large files. This is fixed in extension version 0.1.14.
Thanks to
@PhilipKung5
for reporting the issue.
1
0
10
@jullerino
Fixed
VSC extension version 0.1.47 improves trimming of redundant lines in situations with trailing commas
1
0
13
0
0
8
Join our Discord to get updates on Supermaven and raise issues or feature requests:
0
1
8
VSC extension version 0.1.48 fixes an issue where Supermaven wouldn't return completions if large files were recently committed to the repository
0
1
7
A common evaluation for a long-context model is the "needle in the haystack", where a piece of text (the "needle") is hidden within a large context (the "haystack"). The model is tested on whether it can retrieve the needle after consuming the entire context.
1
0
7
@lino_levan
It's for easy deployment and cross-platform support - most of the complicated logic is in the backend, which is all Rust.
0
0
5
@MenchoRiesco
We would love to support chat in Vim, but our current implementation of chat is a JS app, so it would require a full UI rewrite to support Vim
2
0
5
VSC extension version 0.1.35 fixes an issue where too much whitespace would be deleted before the cursor in some cases.
0
0
5
But the needle in the haystack task is made easier because the needle stands out from the rest of the text: to get it correct, the model only needs to remember the needle.
To give a harder test, we made a "dense retrieval" benchmark.
1
0
4
In this task, the model is given a sequence of key-value pairs. Each pair occurs twice. The model is scored on whether it can use the first occurrence of the pair to correctly guess the value when it sees the pair a second time.
2
0
4
Here, we plot the dense retrieval performance as a function of the number of tokens separating the first and second occurrence of a pair.
Many groups have found models are best at remembering the start and end of the context, and we reproduce this finding.
1
0
3
@1baga
It’s in JetBrains review - we submitted a new version today that should resolve the last remaining issues
0
0
3
@ITLSR
@cursor_ai
We are planning to add a chat feature in the future.
Note Cursor and Supermaven are fully compatible since Cursor is a VSCode fork. You'll just want to disable Copilot++ so they don't conflict
2
0
3
@lino_levan
@zeddotdev
We're not actively developing a Zed plugin at the moment, but it's something we'd love to do
0
0
2
Lastly, we plot prediction error as a function of token index on our internal code, averaging over several hundred orderings of the files to get a smooth result.
1
0
3
@oggiesutrisna
@t3dotgg
We have a Jetbrains extension. It doesn't have this specific fix yet but we will port the fix this week.
1
0
3
@0xwendel
It's just a better scheduler that avoids starving requests. The model is the same, p50/90 latency is unchanged.
0
0
2
