At its user conference in San Francisco this week, GitHub introduced a new AI tool designed to let developers build applications in natural language. Called GitHub Spark, the AI tool creates micro-apps called sparks, which can integrate AI features and external data sources without requiring any management of cloud resources.
Spark uses a feedback loop that starts with an initial prompt, using both OpenAI and Anthropic models. That process generates a live preview of the app as it’s built, and saves versions of each iteration so that developers can compare. It also allows developers to make changes directly to the underlying code.
“Once a user is happy with their spark, they can automatically run it on their desktop, tablet or mobile device, ultimately getting immediate value from their own creation,” the company said in a prepared statement. “They can also share their sparks with customized access control, as well as allow others to remix their spark and build upon their creations.”
Spark is currently in technical preview with a waitlist.
The company also announced support for multiple models from within GitHub Copilot. The models supported are:
- Anthropic’s Claude 3.6 Sonnet
- Google’s Gemini 1.5 Pro
- OpenAI’s GPT-4o, o1-preview and o1-mini
GitHub also announced a slew of new developments for its Copilot tool, including:
- Multi-file edit for GitHub Copilot in VS Code: Users can use Copilot Chat in VS Code to make edits across multiple files at the same time. In this new editing mode, Copilot implements complex changes across a variety of files within a project based on natural language prompts.
- GitHub Copilot Extensions for all users, which allow developers to ask questions of any integrated developer tool, such as leading developer tools and services like Atlassian, Rovo, Docker, Sentry and Stack Overflow. Users will also be able to build their own private extensions that work with their internal developer tooling, the company noted. This will be generally available early next year.
- GitHub Copilot for Xcode: The code completion capabilities of Copilot are now available in public preview for Xcode, empowering developers building apps across all Apple platforms.
- A new code review capability that enables Copilot to offer feedback on code in 30 seconds. Users can request a review from Copilot in Visual Studio Code, or on GitHub.com when they create a pull request.
- Security campaigns, which is currently available in public preview. This uses Copilot Autofix to help developers and security teams prioritize and orchestrate vulnerability remediation at unprecedented scale. For organizations facing backlogs of security debt, teams can set up security campaigns to triage up to 1,000 alerts at a time — as well as filter alerts by type, severity, repository and team. The tool provides the context and code suggestions to fix software vulnerabilities through Copilot Autofix.
Stack Overflow Launches New AI Extension for GitHub Copilot
Sixty-one percent of developers spend more than 30 minutes a day searching for solutions to programming problems, according to the 2024 Stack Overflow Developer Survey. A new extension aims to help them find answers faster.
Stack Overflow launched a GitHub Copilot extension this week that lets developers query the code assistant and receive a summarized response from Stack Overflow.
It’s an interesting move — and perhaps a sign of the times — because earlier this year, some users were angered when Stack Overflow announced a similar API partnership with OpenAI that integrated Stack Overflow’s technical content with OpenAI’s ChatGPT.
The new extension does provide attribution to the original Stack Overflow content used to generate the answer.
“This integration ensures that developers remain in their flow state, not needing to leave and search for knowledge, ultimately increasing productivity and speeding up learning,” the company said in its announcement of the new extension.
The Stack Overflow extension is currently available to a limited number of early access users through the GitHub Marketplace. Stack Overflow is also planning to extend the Stack Overflow extension for GitHub Copilot to its Stack Overflow for Teams product in the coming months.
New Instabug Tool Leverages AI To Stop Mobile App Bugs, Crashes
A new tool AI tool from Instabug, an AI-enabled mobile observability platform, uses AI to resolve bugs and crashes for mobile apps. Called SmartResolve 2.0, it uses AI to help discover and fix app crashes and stability problems.
It relies on a proprietary, fine-tuned AI model to analyze crash report data and app source code to pinpoint the root cause of issues accurately, the company said in a press release. It will automatically generate the code required to resolve the issue. Developers can then review and apply the fix with a single click.
Among the features offered by SmartResolve 2.0 are:
- AI-driven crash analysis;
- Code generation; and
- Integration with code repositories to generate pull requests for quick deployment.
It’s currently part of Instabug’s crash reporting, and is in private beta for customers who opt to have their source code shared with the AI model.
The company is hosting a webinar on Nov. 7 for those who’d like to learn more.
Bun Adds Support for Most Upvoted Features, Promise.try
Bun released v1.1.31 with support for HTTP2 server and gRPC support, which it noted were its most upvoted feature requests.
In Bun, node:http2 runs 2.4x faster than in Node v23, the team added in an Oct. 18 blog post. It also means that developers can also use gRPC with packages like @grpc/grpc-js, the team added.
The team had also wanted to include Deno as well, but noted that Deno doesn’t appear to implement server support for the node:http2 module yet, Bun’s team said in a post on the social platform X.
This release also implements Promise.try
, which is similar to Promise.resolve
, except it also works with synchronous functions.
In a boost for security, the release also supports configuring CA certificates for Bun install.
“This is useful when you need to install packages from your company’s private registry, or if you want to use self-signed certificate,” the team wrote.
Other changes in the new release are:
bun build --drop
, which can remove function calls from a JavaScript bundle. “This is useful if you want to remove debug code from your production bundle,” the team stated;bun --drop
, which also works at runtime;Bun.inspect.table()
, which can be used to format tabular data into a string. “It is similar toconsole.table
, except it returns a string rather than printing to the console”;- Iterable SQLite queries. “Bun has a built-in SQLite API that makes it easy to query SQLite databases. Instead of returning an array of rows, you can now return an iterator that yields rows as they are returned from the database”; and
- New APIs that make it easier to work with JavaScript iterators and generators.
The post GitHub Unveils AI Tool for Micro Apps and Improves Copilot appeared first on The New Stack.
Leave a Reply