AI.info
AI Tools Updates — Page 118
What changed in the AI tools this week — dated, from the makers’ own changelogs.
Em inglês
- n8n — n8n@2.41.1
Agents are enabled by default.
- GitHub Copilot — Local sandboxing in the GitHub Copilot app
- GitHub Copilot — More ways to request and configure Copilot code reviews
- ElevenLabs — Image & Video
Sora 2 and Sora 2 Pro have been removed from the model picker; Seedance 1.5 Pro is no longer offered for new generations.
- Cursor — Rollouts and Security Review
Rollouts watches every change as it deploys and reports its health per environment; Security Review reports bugs on every pull request.
- XO — DS-10 Drum Shaper
Fixed bug in tool tip. Now shows correct percentage range.
- XO — DB-30 Drum Butter
Fixed a bug where the formant-shifter to not handle silence consistently; Fixed a rare bug where sounds sometimes were different when GUI was closed or open
- GLM Coding Plan — 3.14.3 Released Sep 22, 2026
The concurrency limit of a running workflow can now be adjusted directly, without stopping the task.
- SenseChat — SenseNova U1.5 Technical Report: Toward Native Unified Visual Intelligence
SenseNova U1.5 understands, reasons about, and generates visual content without external vision encoders or variational autoencoders.
- RunPod Pods — New Pod deploy flow is now the default
The redesigned Pod deploy flow, where you configure your workload before choosing a GPU, is now the default for all users.
- Google AI Studio — Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS generally available (GA) : Released
Released text-to-speech (TTS) audio models and the Gemini API Voices endpoint (/v1beta/voices).
- Convai — Convai Scene Tagger: Tag an Entire Unreal Engine Level for AI Characters in Minutes
Tag an Unreal Engine level for AI characters.
- Zilliz Cloud — 🐦 Serverless and Free clusters are now compatible with Milvus v3.0.x — Zilliz Cloud has upgraded Serverless clusters to Milvus v3.0.x, so you can use the following features with no action on your sid
Serverless and Free clusters are now compatible with Milvus v3.0.x, with no action on your side.
- Google AI Studio — Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS generally available (GA) : Released our next-generation text-to-speech (TTS) audio models and the Gemini API Voices endpoint ( /v1beta/voices ):
Released Gemini 3.8 Flash TTS, Gemini 3.8 Flash-Lite TTS, and the Gemini API Voices endpoint ( /v1beta/voices ).
- AssemblyAI — Dashboard Settings Moved to Left Navigation
Dashboard Settings, including links and theme settings, moved from the top-right of the navigation to the lower-left.
- Rask AI — Faster Editing: Undo, Keyboard Shortcuts and Playback Speed
Undo, Keyboard Shortcuts and Playback Speed
- Julius AI — Claude Opus 5.5 is Available
Claude Opus 5.5 replaces Claude Opus 5 in Julius.
- Julius AI — Grok 4.7 is Available
Grok 4.7 is now available in Julius, replacing Grok 4.6.
- DeepEval — typescript-v0.9.18
- Swapcard — Booth Roster Cleanup: Remove Members Without Filing a Ticket 🧹
Exhibitor admins can remove booth members, and any booth member can leave the roster.
- Unstructured — 0.27.8
Fixed DOCX and HTML partitioning, list recognition, table headers, document conversion, and Windows process handling.
- GLM Coding Plan — Release v3.14.3
The concurrency limit of a running workflow can now be adjusted directly, without stopping the task.
- GLM Coding Plan — Release v3.14.1
Draft references are now inserted correctly after selecting a plugin.
- UiPath Studio — Installing the GitHub Actions extension now also installs the git extension that it depends on.
Installing the GitHub Actions extension also installs the git extension.
- Consensus — Research Gaps Matrix updates
Users can click into each cell to see papers behind it, review supporting evidence, and explore where research exists or gaps remain.
- Ambience — Introducing Ambience Chorus — Part 2: System of Action
Part 2: The System of Action
- Ramp Bill Pay — Ramp Receivables
Ramp drafts invoices with AI, follows up on late payments on your behalf, and matches cash as payments land.
- Aftershoot Culling — How to Reach Your Wedding Gallery Count With Fewer Culling Passes
Cull to Target lets photographers set a target gallery count before culling and weighs coverage and variety alongside photo quality.
- Taskade — v6.256.0: Taskade EVE Media Automations & Live Updates
Taskade EVE can build media automations, lists update live when teammates make changes, and media triggers start automations.
- Darktrace RESPOND — Darktrace Extends Behavioral Security to Help Customers Safely Adopt Enterprise AI with General Availability of Darktrace / SECURE AI™
General availability of Darktrace / SECURE AI™ to help customers safely adopt enterprise AI.
- Ramp Corporate Card — Ramp Receivables
Ramp drafts invoices with AI, follows up on late payments on your behalf, and matches cash as payments land.
- Fireworks AI — New Serverless Training model: GLM 5.3 Flash
GLM 5.3 Flash is now available for LoRA workloads on the shared Serverless Training pool, with up to 200K context and both text and vision inputs.
- UiPath Studio — The arguments dictionary that the Invoke Workflow File activity opens now has a search box.
A matching entry can be selected while the filter is still applied, and clearing the filter brings that entry into view.
- UiPath Studio — With the Compact Design Experience setting enabled, a container activity no longer displays the header of the Sequence that makes up its body.
A container activity no longer displays the header of the Sequence that makes up its body.
- UiPath Studio — Invoke Workflow File validation no longer stops a build or debug session at the first problem it finds.
All argument mismatches are now reported together, and an argument that differs only between In and In/Out is reported as a warning rather than an error.
- UiPath Studio — The agent-ready timeout used when starting a remote debugging session is now configurable, for machines where signing in takes longer than the previous fixed wait allowed.
The agent-ready timeout used when starting a remote debugging session is now configurable.
- Aftershoot Editing — How to Reach Your Wedding Gallery Count With Fewer Culling Passes
Cull to Target lets users set a gallery-wide target photo count before culling and weighs coverage and variety alongside quality.
- Consensus — September 22nd, 2026 - Research Gaps Matrix updates
Users can click into each cell to see the papers behind it, review supporting evidence, and explore where research exists or gaps remain.
- AudioShake — AudioShake Releases Multi-Speaker 2.0 to Isolate Overlapping Dialogue into Discrete Stems
Isolate overlapping dialogue into discrete stems
- Fern SDK Generator — (fix): The generated OAuth wire-test mock ( tests/wire/mockAuth.ts ) now expects the grant_type the generated OAuthAuthProvider actually sends.
The mock expects grant_type: "client_credentials" when the token endpoint models grant_type as an optional string and the example omits it.