AI.info
AI Tools Updates — Page 59
What changed in the AI tools this week — dated, from the makers’ own changelogs.
In inglese
- Replicate — Agent skills for Replicate
Replicate now publishes agent skills, markdown instruction files that give coding assistants knowledge about working with AI models on Replicate.
- AI Dungeon — Voyage Launches the First AI-Native RPG Platform
Voyage launches an AI-native RPG platform.
- LlamaIndex — v0.14.21
Fixed core issues in node deletion, structured output failures, persistence encoding, and message block buffer resolution.
- Zenlytic — Secure Web Search for AI Analytics Agents: Why It Matters and How We Built It
Zenlytic’s new web search feature lets your AI analyst combine real-time web context with private company data.
- Leonardo AI — April 20, 2026 — mode parameter retiring May 4, 2026
The mode parameter for GPT Image-1.5 and Ideogram 3.0 generations is being retired on May 4, 2026. Use the quality parameter instead.
- Ninja AI — Introducing Claude Opus 4.7 on SuperNinja
Claude Opus 4.7 is now live in SuperNinja's Complex Plus mode.
- ResearchRabbit — Minor release 2026-04-17
Minor release
- Descript — Release roundup—April 16th, 2026
Transitions now come with animated previews and are organized into a new grid layout with thumbnail previews.
- Character.ai — introducing c.ai books: classic literature, now playable
Character.ai launched c.ai Books, where users can interact with classic literature on mobile and web through c.ai Labs.
- Norm Technology — Norm Ai Launches The Legal AGI Lab to Build the Legal Infrastructure for AI Agents
Build the legal infrastructure for AI agents.
- Creatify — Introducing AdFlow: the node-based ad builder built for production scale
Node-based ad builder built for production scale
- Gensyn Testnet — Introducing AXL: Peer-to-Peer Communication for AI
- Dataiku LLM Mesh — New feature: Leverage existing external Vector Stores
Leverage existing external Vector Stores
- DeepSource — Upgrades to AI Review Engine
AI Review has new underlying models, improved analysis pipelines, two review tiers, mention-triggered runs, and processed LOC billing.
- Marble — Streaming 3DGS worlds on the web
Describes Spark 2.0's system for 3D Gaussian Splatting.
- QA Wolf — Test iOS Apps Behind a VPN
QA Wolf supports VPN configuration for iOS app testing, with sessions configured per test run and torn down automatically.
- Browser Use Cloud — BYOK, Code Mode & Sensitive Data
Use your own LLM API keys with V3 sessions; the agent writes you a reusable Python script; pass secrets to the agent securely.
- DeepSource — Configure New Vulnerability Alerts
Configure email recipients, organization-admin notifications, and the minimum severity level for new vulnerability alerts.
- Empathic Voice Interface (EVI) — EVI API additions
Added configurable turn detection and interruption settings to EVI configs.
- Pika — Your Pika Agent Just Got a Job
We're making it possible for your Pika Agent to earn real money — every conversation, every skill used, every interaction counts.
- Kagi Assistant — Tuning the Orchestra
Assistant changes addressed mobile app display, page context, thread dates, research searches, SSH commands, and editing prompts with files.
- Kagi Search — Tuning the Orchestra
Small Web adds a tap-based way to switch posts; Translate adds wordplay detection, language options, and keyboard shortcut improvements.
- Omi — Phone Calls, Bluetooth Overhaul, Offline Sync & More
Omi can make phone calls; action items sync with Apple Reminders; offline recordings sync in the background.
- Luminar Neo — Luminar Neo 1.27.0: Smarter portrait tools, mask feathering and more
Adds Bokeh AI, new Skin AI controls, Dark Circles removal in Face AI, and the Mask Feather control.
- Cerence CaLLM — Cerence AI to Power Intelligent, LLM-Powered In-Car Experiences for BYD
Cerence AI will power intelligent, LLM-powered in-car experiences for BYD.
- Genspark — Introducing Genspark AI Workspace 4.0: Your AI Employee, Now Everywhere
Introducing Genspark AI Workspace.
- Enterprise h2oGPTe — v1.7.0
Introduces Human-in-the-Loop agent control, Memory Blocks, a native MCP server, scheduled tasks, and governance features.
- DeepSource — DeepSource MCP Server
The MCP Server exposes 30 tools across 8 categories for code review findings, PR grades, vulnerabilities, metrics, compliance reports, and issue suppression.
- MLflow — MLflow 3.11.1 Highlights: Automatic Issue Detection, Gateway Budget Management, and Pickle-Free Models!
Detect issues in traces, set Gateway budget policies, visualize trace graphs, export OTel GenAI conventions, trace OpenCode, infer UV dependencies, and serialize models…
- Reducto — Introducing Deep Extract
We're releasing Deep Extract, an update to structured extraction.
- Labelbox Model Foundry — April 6, 2026
Custom roles compatible filters; audio transcription in audio & video editors.
- QA Wolf — Test on Any Android Device and OS. Tablets Too.
Access any Android device and OS combo, with infrastructure to run tests in parallel.
- Cohere — Retirement of Embed v2.0 and Aya Expanse / Vision 8B
Effective April 4, 2026, the listed models are no longer available; requests using these model IDs will fail.
- OLMo — v2.5.0
Added an exponential LR scheduler, Peri-LN transformer block, and support for loading and averaging multiple checkpoints.
- Browser Use Cloud — Free Tier & New Pricing
Browser Use Cloud is now free. Proxy data cost cut from $10/GB to $5/GB; custom proxies now available on all subscription plans.
- LlamaIndex — v0.14.20
- Writesonic — Writesonic GEO (AI Visibility) Data Now Available in Google Looker Studio
Connect your AI visibility data to Looker Studio. Build custom dashboards alongside GA, Search Console, and paid media.
- AnythingLLM — AnythingLLM v1.12.0
Added automatic native tool calling, Intelligent Tool Selection, and a Filesystem Agent.
- Graphite — Introducing Code Tours: a new way to review
Introduced Code Tours as a new way to review.
- DeepSource — Enterprise Server v5.0.0
Enterprise Server customers can run AI Review using their own model provider credentials; inference calls go directly to their chosen provider.