LogoValueSpot
FeaturesUse CasesPricingGuidesCommunity
Get Started

© 2024 ValueSpot. All Rights Reserved.

Privacy PolicyTerms of Service

    News History of https://nitter.privacydev.net/amanrsanger/rss

    1.
    So excited to be working with @jbfja and the Supermaven team! There are only two companies in the world that have shipped models with 1M+ token windows: Google Deepmind and Supermaven.

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-11-21 12:50:59 AM
    2.
    R to @amanrsanger: Can't wait to show everyone what we've been working on!

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-11-21 12:50:59 AM
    3.
    The Benefits of Pair Programming in an IDE

    The post discusses the experience of coding in an Integrated Development Environment (IDE) and the benefits of pair programming, including having a trusted partner to learn from and a tabbed workspace.

    Source: https://nitter.privacydev.net/amanrsanger/rss

    AIIDE
    		Rating: 61
    2024-11-04 10:50:49 AM
    4.
    R to @amanrsanger: But, there are regions of code/edits where your intent just lives in your head. With minimal effort, you should be able to inject this intent into Cursor. (As a toy example, just typing inline instructions informs Cursor to jump to the correct location) [3/7]

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    5.
    R to @amanrsanger: Furthermore, the model shouldn't just increase your words per minute. Can we actively surface the next most helpful bits of information for finishing a change? This could be a piece of code, docs, or a helpful comment/suggestion. [4/7]

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    6.
    R to @amanrsanger: The ceiling of suggestion quality is much higher than today! We can draw on more intelligent models to propose larger/harder actions. Then fall back on smaller models for easier ones. And if the next action is indeterminate, we propose nothing! [5/7]

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    7.
    R to @amanrsanger: The result is a model that stays a few steps ahead of the programmer. Always jumping them forward in time, always preserving their flow. [6/7]

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    8.
    R to @amanrsanger: To solve this, we’ll need to do the following: - Fundamental research on action prediction across a codebase - Continued pre-training and post-training on ~5-13B active parameter code-models - Additional inference tricks similar to spec edits (https://cursor.sh/blog/instant-apply) [7/7]

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    9.
    R to @amanrsanger: Feel free to reach out/dm me if next action prediction or any of these problems seem interesting to you! And read more abt the other problems here! https://anysphere.inc/blog/problems-2024

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    10.
    Priompt is one of the few prompting libraries that passes the sniff test for large-scale production applications. Even Copilot uses it!

    Source: https://nitter.privacydev.net/amanrsanger/rss

    2024-10-30 08:10:47 AM
    1
    23