xAI Launches Grok-Imagine-Image: A Dual-Mode AI Image Generation and Editing Model

·1 views

xAI has released Grok-Imagine-Image, an AI model available on Replicate that supports both text-to-image generation and image-to-image editing. Users can supply a text prompt alone to generate new images, or pair it with an existing image to guide edits such as background replacement, style changes, or composition refinement. The model accepts standard formats including JPG, PNG, and WebP, and outputs a single image URI across multiple aspect ratios. A key limitation is that when an input image is provided, the model switches to editing mode and ignores the aspect ratio parameter, with output dimensions determined by the source image instead. Practical use cases include product photography enhancement, marketing asset iteration, concept art refinement, and interior design visualization.

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)

Fluentic Style's createSheet Solves Third-Party DOM Styling Without Owning Elements

Fluentic Style, a CSS-in-JS styling library, introduced a utility called createSheet to address a common front-end challenge: styling DOM elements rendered by third-party libraries that developers do not directly control. Components like React Select, date pickers, and rich text editors generate their own internal markup with fixed class names, making direct element-level styling impossible for the consuming developer. createSheet works by attaching a group of selector-based styles to a single boundary element, so that when a generated class is applied at the wrapper level, it cascades styles down to known internal selectors. Developers first map third-party class names to named style targets using style.selector(), then group those targets into a sheet that is applied at the component boundary. This approach consolidates third-party selector references in one place rather than scattering them across CSS files, keeping styling logic structured and maintainable.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Built an Autonomous AI Agent to Apply for Jobs Every 4 Hours Overnight

A software developer configured Grok Bot to act as an autonomous job-search agent, tasking it with finding, evaluating, and applying to software engineering roles on a recurring four-hour schedule while he slept. Rather than maximising application volume, the agent was designed with a scoring system to prioritise interview probability, applying only to roles that closely matched his resume. The developer wrote a detailed operating policy as the prompt, instructing the bot never to fabricate credentials and to skip weak matches entirely. Grok Bot's built-in environment — supporting browser automation, memory, reusable skills, and scheduled routines — reduced the need to build custom infrastructure from scratch. The experiment highlights a broader shift from using AI as a conversational tool to delegating it autonomous, multi-step work tasks.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

ToolSuture Fixes Broken AI Agent Tools Without Touching the Agent's Code

A developer built ToolSuture, an AI recovery system, as an entry for the All Things Agentic Hackathon to address a overlooked reliability problem: tools changing underneath a deployed agent. Rather than rewriting the agent when an external API or tool contract changes, ToolSuture runs an observe-diagnose-repair loop using Gemini 1.5 Flash on Vertex AI to assess whether the old and new tool contracts remain semantically equivalent. If they do, the system creates a compatibility layer around the frozen agent and replays the original mission to verify success — with zero changes to the agent's code. Crucially, ToolSuture also refuses to act when a change appears structurally similar but is operationally different, such as a recoverable deletion becoming permanent — blocking execution entirely. The project reframes agent autonomy not as permitting more actions, but as distinguishing safe recoverable changes from dangerous irreversible ones.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

pg_igraph Brings Cypher-Like Graph Queries Inside PostgreSQL With Its Own Grammar

The pg_igraph extension allows developers to run graph traversal queries directly within PostgreSQL using a Cypher-inspired MATCH syntax, but it operates through its own lexer, parser, and AST rather than simple string substitution. A shorthand PATH FROM...TO...VIA form is also available for straightforward point-to-point path lookups without writing full MATCH expressions. The extension supports JSON-backed query parameters, keeping application values separate from the graph expression itself and enabling cleaner, scalable code. Notably, row-returning MATCH queries and mutation or path commands return different response shapes, which developers must account for in application logic. Because the content inside igraph_query() is not standard SQL, tools like ORMs, linters, and query proxies may treat it as an opaque argument, making the abstraction boundary between PostgreSQL and the graph layer an important design consideration.

0 comments Read more at DEV Community