Agents coordinate code changes across hundreds of repositories
Agentic Batch Changes automates multi-repo migrations by writing scripts when possible and delegating judgment calls to Claude/Codex, replacing manual spreadsheet tracking and oversized PRs.
Eliminates the coordination tax of managing hundreds of PRs across heterogeneous repos—what used to require manual tracking now tracks merge status from a single view. Reduces the cognitive load of validating changes landed everywhere they needed to.
Replaces spreadsheet-based migration tracking and manual scripted changes. Requires Sourcegraph Cloud indexing of your codebase and integration with your VCS. Ready now for orgs already using Sourcegraph; outcome-based pricing (pay per merged changeset) removes token-cost uncertainty. Start with security fixes or library migrations.
- “Agentic Batch Changes writes scripts more often than it deploys coding agents, which is what keeps it efficient at scale”
- “Customers have merged hundreds of thousands of changesets this way, tens of millions of lines of code”
- “The largest single Batch Change we have seen merged more than 2,200 changesets from one change”
- “you pay per changeset merged into your codebase, not per token, seat, or attempt”
- “During the Beta, customers merged nearly a thousand changesets created by Agentic Batch Changes”
batch-changesmulti-repoagentic-codingsourcegraph
Ollama adds decision models for classification tasks
Ollama's /v1/systemone endpoint returns probabilities and confidence scores instead of text, replacing multi-turn prompts for ticket triage and routing with single structured requests.
Eliminates prompt engineering overhead for classification workflows by returning calibrated probabilities across predefined categories. Reduces latency and token overhead compared to parsing text responses for deterministic decisions.
Replaces custom parsing logic around text-based classification. Requires Ollama v0.35.0+, one of two available models (Nimble or Tev1), and reformatted requests using the /v1/systemone schema. Ready to try now for ticket triage and content routing, but model coverage is minimal—validate Nimble/Tev1 fit your classes first.
- “Ollama now supports decision models through /v1/systemone”
- “Decision models return choices, probabilities, and scores instead of text”
- “Available models: Nimble from Bespoke Labs, Tev1 from Together AI”
- “The API supports three question types: choice, noul, score”
classificationollamastructured-outputticket-routinglocal-inference
EuroLLM and Apertus join Workers AI platform
Two European open models (EuroLLM: 35 languages; Apertus: 1,500 languages) now available via Workers AI for sovereign, multi-language inference without vendor lock-in.
Developers can now build multilingual applications on non-US infrastructure without frontier model dependencies. Apertus notably supports rare/regional languages (Romansh, Swiss German, low-resource Asian/African languages) where closed models underperform, unlocking new use cases for underserved regions.
Replaces dependency on closed frontier models for EU/multilingual workloads. Requires requesting access via Workers AI platform; no code changes if already using Workers AI inference APIs. Ready now—both models published with full weights/training data. Apertus explicitly designed for GDPR/EU AI Act compliance, reducing legal friction in regulated environments.
- “EuroLLM supports 35 languages, including all 24 official EU languages”
- “Apertus was trained on more than 15 trillion tokens across more than 1,500 languages, with 40% of training data in languages other than English”
- “Apertus significantly outperforms leading closed and open models on rare and regional languages, from Romansh and Swiss German to low-resource languages across Asia and Africa”
- “Its architecture, weights, training data and methods are all published”
- “built by public institutions, for the public good”
open-modelsmultilingualworkers-aisovereigntyinfrastructure
DeepSeek V4 Flash Vision now experimental on AI Gateway
Multimodal vision model (JPEG/PNG/GIF/WebP) integrated into Vercel's AI Gateway with tool use, reasoning, and caching parity to text-only version.
Adds image understanding to fast inference requests without switching providers or managing separate vision endpoints. Developers building screenshot analysis, chart parsing, or vision agents can consolidate model selection.
Drop-in replacement for text-only DeepSeek V4 Flash if vision input exists; requires no new dependencies beyond `ai` SDK. Experimental flag (`-exp`) means API surface may shift—configure fallback for production paths. Worth prototyping now for latency-sensitive vision tasks.
- “DeepSeek V4 Flash with vision”
- “accepts images alongside text”
- “Images can be JPEG, PNG, GIF, or WebP”
- “Tool use, reasoning, and caching all work the same as before”
- “The `-exp` in the model ID marks this as an experimental release. Expect behavior to change”
- “model: 'deepseek/deepseek-v4-flash-vision-exp'”
- “AI Gateway reflects provider pricing with no markup and does not charge a platform fee on inference”
vision-modelsmultimodalai-gatewaydeepseekexperimental
Omni 1.1 Flash adds scene extension, 4K upscaling
Gemini Omni 1.1 now analyzes 10 seconds of prior context for seamless video extension, supports keyframe interpolation, and generates 360p previews 60% faster—available via Gemini API with production-ready controls.
Cuts iteration time for video generation workflows through faster 360p drafting and deterministic output via keyframe specification. Enables longer narrative continuity in generative video without flickering or context loss.
Replaces ad-hoc frame stitching and manual keyframe hunting. Requires Gemini API key and updates to prompt structure (`previous_interaction_id` for extensions, video references up to 3 seconds). Ready now—rolling out across Google AI Studio, Enterprise Agent Platform, and subscriber tiers.
- “the model can now analyze up to 10 seconds of prior context — a leap from previous models that only referenced the final second”
- “Generate lightweight previews in 360p resolution up to 60% faster and at a third of the cost compared to Omni 1.1's standard 720p resolution”
- “You can extend videos in 10-second increments up to a total cumulative length of 40 seconds”
- “Reference up to three seconds of video when crafting your scene”
video-generationgemini-apigenerative-mediaapi-releaseproduction-ready