AWS Launches Nova 2 Model Family With Lite, Pro, and Omni Tiers on Bedrock

·1 views

Amazon Web Services has released the second generation of its Nova foundation model family, available through Amazon Bedrock. The lineup comprises three models — Nova 2 Lite, Nova 2 Pro, and Nova 2 Omni — each targeting different workloads across a cost-versus-capability spectrum. Nova 2 Lite is optimized for high-volume, lower-cost tasks such as chatbots and document processing, while Nova 2 Pro offers a 1-million-token context window suited for complex reasoning and large enterprise workflows. Nova 2 Omni is designed for advanced any-to-any multimodal processing spanning text, images, video, and speech. Key improvements over the original Nova family include extended thinking controls, broader multilingual support covering over 200 languages, higher throughput, and lower inference costs.

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)

Why ATS Keyword Checkers Flag Skills You Already Have — And How to Respond

Many popular ATS keyword checker tools flag resume skills as 'missing' because they perform simple string matching rather than understanding language variations like synonyms, plurals, or word order differences. For example, writing 'analyzed sales data' may be flagged as lacking 'data analysis' even though the meaning is identical. Actual recruitment platforms such as Workday, Greenhouse, and Lever are generally more sophisticated than these third-party checkers, making many of their warnings false alarms. Experts advise job seekers to check whether the underlying skill or concept is present rather than obsessing over exact phrasing, and to align wording with the job posting only once where genuinely applicable. The ultimate priority should be writing a resume that reads naturally for human recruiters, since a keyword-stuffed document may clear automated filters but fail to impress the person making the hiring decision.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Ada Lovelace AI Tool to Generate Personalized Code Compliments

Developer Mike Cramblett has launched Ada Loves Code (ALC), an open-source tool that uses an Augusta Ada Lovelace persona powered by AI to inspect code repositories and generate personalized "Certificates of Poetical Science." The project, remixed from his earlier hackathon entry HypeBESTIE using Google AI Studio, aims to provide developers with meaningful, bespoke recognition beyond generic praise. Cramblett was motivated by his own experience of sharing work online for over a year with little engagement, and wanted to address the silence many developers face after publishing projects. To demonstrate the tool, he browsed recent weekend challenge posts, identified low-visibility entries, and used ALC to generate and post personalized certificates directly in their comment sections. The initiative reflects his broader belief that developer communities can cultivate a more generous, substantive peer-review culture online.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Voice-Controlled AI Coding Tool to Work While Walking Outside

Software developer Benoit Hirbec created Squad33, a macOS app that lets users control AI coding agents entirely by voice, after chronic back pain in March 2026 forced him to find an alternative to desk-based work. The tool uses Whisper for transcription on a Mac, routing spoken instructions to Claude Code or Opencode, with responses read aloud through earbuds on a paired phone. This setup allows Hirbec to dictate tasks during outdoor walks, with AI agents handling development work in parallel git worktrees while he moves on to the next feature. He reports that listening to agent output is less mentally draining than reading it at a screen, and that voice input removes the keyboard fatigue that often limits detail in written prompts. Squad33 is currently available as a free macOS beta and works with existing Claude Code or Opencode subscriptions rather than providing its own model access.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Bug Fix Stops Internal Code Fragments From Leaking Into GitHub Issue Titles

A recurring defect in an automated task system caused raw internal checklist and prompt scaffolding — such as fragments like '#2637/brand#16' — to appear as GitHub issue and pull request titles instead of meaningful human-readable labels. The bug originated in a function called make_issue_title within tasks/issue_title_builder.py, where boundary logic validated only that a string was present and the API call succeeded, not that the resulting title was useful to a human reviewer. Because API requests completed without errors, the failure produced no crashes or alerts, making it easy to overlook during normal development workflows. The fix restructures the boundary logic to source titles from the task's description field — the canonical human-facing label — and flags any checklist- or prompt-shaped strings as leakage even when they are technically API-valid. New assertions in tests/test_title_generation.py now detect known leakage patterns and emit warnings, ensuring internal scaffolding no longer silently passes into public-facing artifacts.

0 comments Read more at DEV Community