OpenAI and Broadcom Unveil Jalapeño, a Custom AI Inference Chip

·1 views

OpenAI and Broadcom jointly announced Jalapeño on June 24, 2026, marking OpenAI's first custom Application-Specific Integrated Circuit (ASIC) designed exclusively for large language model inference. The chip pairs a reticle-sized compute chiplet with High-Bandwidth Memory, placing storage physically close to processing units to reduce the data-movement delays that typically bottleneck token generation. Unlike general-purpose GPUs, which handle training, graphics, and diverse workloads, Jalapeño is hardwired solely for inference, trading versatility for greater energy efficiency. Early testing indicates substantially improved performance-per-watt compared to GPUs, though final figures are still being measured. The design went from conception to tape-out in approximately nine months, which OpenAI describes as one of the fastest chip development cycles it has undertaken.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Expo Raises $45M Series B to Accelerate App Development Tools and AI Features

Expo, the mobile app development platform, has raised $45 million in a Series B funding round led by Georgian, which has previously backed developer infrastructure companies such as Replit and Render. Co-founder Charlie Cheever noted that Expo was already profitable and did not need the capital to sustain operations, but chose to raise funds to speed up product development. The company plans to use the investment to hire engineers and build new features that make app creation faster and more accessible. Growing adoption of Expo alongside AI tools has expanded the team's roadmap significantly. Expo also stated a broader ambition to enable non-developers with vision and determination to build application software.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Static Scanner to Catch Supply-Chain Malware Before Code Runs

A developer created a static repository scanner after being targeted by a fake job interview repo designed to steal credentials and API keys. The tool analyzes repositories without cloning, installing, or executing any code, since malicious scripts embedded in npm lifecycle hooks like postinstall can run automatically during npm install. The scanner checks for three key threat categories: dangerous build-time execution hooks, dependencies in package.json that resolve to unexpected or non-registry URLs in the lockfile, and obfuscation patterns such as eval of decoded strings or long hex literals. The attack that inspired the tool exploited a mismatch between a trusted-looking package name and a malicious tarball URL hidden in the lockfile. By keeping analysis logic free of any I/O and testable against known-malicious fixtures, the scanner can flag suspicious combinations of signals before any harm is done.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Releases 9 Open-Source Telegram Bot Templates Built in Python

A developer has published nine production-ready Telegram bot templates on GitHub, covering use cases such as AI assistance, appointment booking, crypto price alerts, job listings, and survey collection. The bots are built with Python 3.12 and the aiogram framework, using SQLite for data storage and Anthropic's Claude Haiku model for AI features. A freemium model is implemented across relevant bots, allowing 20 free messages per day with unlimited access for premium users. Supporting tools include APScheduler for scheduled tasks and free APIs such as CoinGecko for cryptocurrency data. All nine templates are freely available on GitHub, and the developer is also offering custom bot development starting at $49 per bot.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Build Real-World API Tests Using Python and Pytest

As software systems grow increasingly dependent on microservices and third-party integrations, automated API testing has become essential for maintaining reliability and security. Manual tools like Postman are useful for exploration but fall short in CI/CD environments with hundreds of endpoints. Python's Pytest framework, combined with the Requests library, offers a scalable solution for automating API validation. Developers can write test cases to verify correct status codes, data structures, and error handling — such as confirming a valid user returns HTTP 200 while a missing one returns 404. Advanced patterns like parameterized testing further extend Pytest's usefulness in production-grade testing pipelines.

0 comments Read more at DEV Community