# forck.live — 13–19 April 2026

> AI lab announcements, news reports and podcasts.

forck.live follows what companies announce about their own products. Press may
corroborate an official story; it never leads and never appears as an item of its
own. A missing date stays missing. A title is the publisher's words — where a
company does not publish in English, an English rendering is listed separately
as `Title in English` and named as this site's translation, never as theirs.

This file is one issue of the weekly archive: every confirmed
first-party announcement of ISO week 2026-W16, 13–19 April 2026, newest
first. The window never moves, so this URL always names the same seven days —
but a post published inside them and detected later is added when it arrives,
so a closed week can still gain entries.

Every entry is one first-party announcement, and carries the same eight fields
in this order before any others:

    Company · Announced · Category · Coverage · Announcement · Group · Source · Record

`Source` is the company's own page the announcement was traced to and `Record`
is this site's page for it. Both are safe to cite.

`Coverage` is a number of outlets, or `not counted` — which means this record
reached no count of its own, never that nobody wrote about it.

`Category` is this tracker's own analysis, and is authoritative wherever it is
stated. `not stated` means either that nothing has analysed the post yet or
that the analysis placed it outside this taxonomy; neither is a verdict about
the announcement.

`Announcement` is whether the post is news of something the company did, rather
than a how-to, a customer story, a changelog listing or a staffing note.

`Group` is which of this record's groups the entry falls in, carried on the
entry itself so it holds where the groups are headings and where they are not.
One of: `models`, `covered`, `announcements`, `routine`.

Anything after `Record` is held detail that only some announcements carry.

- This week's page: https://forck.live/week/2026-W16
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Sakana AI — Digital Ecosystems: Interactive Multi-Agent Neural Cellular Automata

- Company: Sakana AI (sakana.ai)
- Announced: 2026-04-18T15:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://sakana.ai/digital-ecosystem
- Record: https://forck.live/items/4690-digital-ecosystems-interactive-multi-agent-neural-cellular-automata
- Subject: Sakana models

Sakana AI releases Digital Ecosystems, a browser-based platform for interactive artificial life research where competing neural networks learn and adapt in a shared 2D grid.

### Alibaba — Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

- Company: Alibaba (alibaba.com)
- Announced: 2026-04-18T02:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen3.6-max-preview
- Record: https://forck.live/items/4603-qwen3-6-max-preview-smarter-sharper-still-evolving
- Subject: Qwen

Alibaba's Qwen team releases Qwen3.6-Max-Preview, an early preview of their next proprietary model with improved agentic coding, world knowledge, and instruction following over Qwen3.6-Plus. Available via Alibaba Cloud Model Studio API and Qwen Studio.

### Lovable — Claude Opus 4.7 now in Lovable

- Company: Lovable (lovable.dev)
- Announced: 2026-04-17T18:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lovable.dev/blog/opus-4-7-now-in-lovable
- Record: https://forck.live/items/12089-claude-opus-4-7-now-in-lovable
- Subject: Lovable / AI app builder

Lovable has added Claude Opus 4.7 to its platform. For a limited time until April 30, user credits go up to 2x further. Lovable's internal benchmarks show Opus 4.7 solves tasks in 40% fewer turns, uses 10–20% fewer tokens, is 15% faster, and scores 2–3% higher on performance compared to Opus 4.6.

### Lovable — What CTOs ask before adopting AI development tools (and how Lovable thinks about the answers)

- Company: Lovable (lovable.dev)
- Announced: 2026-04-17T17:09:03+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/what-ctos-ask-before-adopting-ai-development-tools-and-how-lovable-thinks-about-the-answers
- Record: https://forck.live/items/12090-what-ctos-ask-before-adopting-ai-development-tools-and-how-lovable-thinks
- Subject: Lovable / AI app builder

Lovable published a guide addressing five operational concerns that CTOs and CIOs raise when evaluating AI development tools, covering shadow IT governance, vendor evaluation, security controls, specification clarity, and team adoption.

### Vercel — How Zo Computer improved AI reliability 20x on Vercel

- Company: Vercel (vercel.com)
- Announced: 2026-04-17T04:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/how-zo-computer-improved-ai-reliability-20x-on-vercel
- Record: https://forck.live/items/20716-how-zo-computer-improved-ai-reliability-20x-on-vercel
- Subject: v0 / AI SDK / AI Gateway

Zo Computer, a personal AI cloud provider, reported a 20x reduction in retry rate (from 7.5% to 0.34%) and a 38% cut in P99 latency (from 131s to 81s) after switching to Vercel's AI SDK and AI Gateway. The company also achieved a 99.93% chat success rate, up from 98%. New model support was reduced from an hour-long code change to adding a config string in 30 seconds.

### xAI — Grok Speech to Text and Text to Speech APIs

- Company: xAI (x.ai)
- Announced: 2026-04-17
- Category: api-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://x.ai/news/grok-stt-and-tts-apis
- Record: https://forck.live/items/3473-grok-speech-to-text-and-text-to-speech-apis
- Subject: Grok

xAI announces Grok Speech to Text and Text to Speech APIs, highlighting fast and accurate performance, natural and expressive voices, simple pricing, and multilingual support.

### Google Research — Designing synthetic datasets for the real world: Mechanism design and reasoning from first principles

- Company: Google Research (research.google)
- Announced: 2026-04-16T14:41:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/designing-synthetic-datasets-for-the-real-world-mechanism-design-and-reasoning-from-first-principles/
- Record: https://forck.live/items/1345-designing-synthetic-datasets-for-the-real-world-mechanism-design-and-reasoning
- Subject: Research

Google Research introduces Simula, a framework that reframes synthetic data generation as dataset-level mechanism design, using reasoning to architect datasets from first principles with fine-grained control over coverage, complexity, and quality.

### Harvey — Opus 4.7, Now Live in Harvey

- Company: Harvey (harvey.ai)
- Announced: 2026-04-16T14:30:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/opus-4-7-now-live-in-harvey
- Record: https://forck.live/items/17582-opus-4-7-now-live-in-harvey
- Subject: Harvey

Harvey made Claude Opus 4.7, Anthropic's latest frontier model, available in its platform. The model scored 90.9% on Harvey's BigLaw Bench evaluation suite, the highest score of any Claude model in Harvey to date, with 45% of tasks receiving perfect scores. Harvey noted improved reasoning calibration and performance on ambiguous tasks, but also some polish gaps relative to Opus 4.6, such as tone mismatch and non-standard citations.

### Google Research — AI-generated synthetic neurons speed up brain mapping

- Company: Google Research (research.google)
- Announced: 2026-04-16T12:18:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/ai-generated-synthetic-neurons-speed-up-brain-mapping/
- Record: https://forck.live/items/1346-ai-generated-synthetic-neurons-speed-up-brain-mapping
- Subject: Research

Google Research introduces MoGen, a model that generates synthetic neuron shapes to improve AI reconstruction models for brain mapping, reducing reconstruction errors by 4.4%.

### OpenAI — Codex for (almost) everything

- Company: OpenAI (openai.com)
- Announced: 2026-04-16T10:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/codex-for-almost-everything
- Record: https://forck.live/items/180-codex-for-almost-everything
- Subject: GPT / ChatGPT / API

OpenAI updated the Codex app for macOS and Windows, adding computer use, in-app browsing, image generation, memory, and plugins to accelerate developer workflows.

### Vercel — A new programming model for durable execution

- Company: Vercel (vercel.com)
- Announced: 2026-04-16T04:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/a-new-programming-model-for-durable-execution
- Record: https://forck.live/items/20717-a-new-programming-model-for-durable-execution
- Subject: v0 / AI SDK / AI Gateway

Vercel Workflows is the easiest way to write long-running, durable, reliable, and observable agents and backends. The gap between prototypes and production-ready systems is huge.

### OpenAI — Introducing GPT-Rosalind for life sciences research

- Company: OpenAI (openai.com)
- Announced: 2026-04-16T01:00:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/introducing-gpt-rosalind
- Record: https://forck.live/items/181-introducing-gpt-rosalind-for-life-sciences-research
- Subject: GPT / ChatGPT / API

OpenAI introduces GPT-Rosalind, a frontier reasoning model for life sciences research, focusing on drug discovery, genomics, and protein reasoning.

### Factory AI — Factory raises $150M Series C

- Company: Factory AI (factory.com)
- Announced: 2026-04-16
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://factory.com/news/series-c
- Record: https://forck.live/items/17871-factory-raises-150m-series-c
- Subject: Droid

Factory AI raised $150 million in Series C funding led by Khosla Ventures, with participation from Sequoia Capital, Blackstone, Insight Partners, Evantic Capital, 20VC, NEA, and Mantis VC, at a $1.5 billion valuation. The company plans to use the capital to accelerate investment in research, product, and global go-to-market. Factory also announced that its Droids are used by hundreds of thousands of developers across enterprises including Nvidia, Adobe, EY, Palo Alto Networks, and Adyen, and that revenue has doubled month over month for each of the past six months.

### Luma — Wonder Project and Luma Launch Innovative Dreams: Production Services Company Combining Generative AI and Traditional Filmmaking, Backed and Supported by Amazon Web Services

- Company: Luma (lumalabs.ai)
- Announced: 2026-04-16
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://lumalabs.ai/news/luma-innovative-dreams
- Record: https://forck.live/items/7966-wonder-project-and-luma-launch-innovative-dreams-production-services-company
- Subject: Ray / Dream Machine

Wonder Project and Luma announced the launch of Innovative Dreams, a production services company that combines generative AI and traditional filmmaking, backed by Amazon Web Services. The company will use a new method called Realtime Hybrid Filmmaking and will provide services to third-party studios.

### Hugging Face — Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-16
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/train-multimodal-sentence-transformers
- Record: https://forck.live/items/1518-training-and-finetuning-multimodal-embedding-reranker-models-with-sentence
- Subject: Platform

This blog post by Tom Aarsen on Hugging Face explains how to train or finetune multimodal embedding and reranker models using the Sentence Transformers library, with a practical example of finetuning Qwen/Qwen3-VL-Embedding-2B for Visual Document Retrieval, achieving improved NDCG@10 from 0.888 to 0.947. It covers the training components: model, dataset, loss function, training arguments, evaluator, and trainer.

### Hugging Face — The PR you would have opened yourself

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-16
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/transformers-to-mlx
- Record: https://forck.live/items/1517-the-pr-you-would-have-opened-yourself
- Subject: Platform

Hugging Face announces a Skill and test harness to help port language models from the transformers library to mlx-lm, aiming to make models available shortly after they land in transformers. The tool is designed to assist both contributors and reviewers by automating scaffolding, architecture porting, and numerical verification.

### Hugging Face — Ecom-RLVE: Adaptive Verifiable Environments for E-Commerce Conversational Agents

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-16
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/ecom-rlve
- Record: https://forck.live/items/1516-ecom-rlve-adaptive-verifiable-environments-for-e-commerce-conversational-agents
- Subject: Platform

Ecom-RLVE, an extension of the RLVE framework to multi-turn, tool-augmented e-commerce conversational agents, introducing eight verifiable environments (product discovery, substitution, cart building, returns, order tracking, policy QA, bundle planning, and multi-intent journeys) with procedural problem generation, a 12-axis difficulty curriculum, and algorithmically verifiable rewards. The authors train a Qwen 3 8B model with DAPO for 300 steps and present early results.

### OpenAI — Accelerating the cyber defense ecosystem that protects us all

- Company: OpenAI (openai.com)
- Announced: 2026-04-16
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/accelerating-cyber-defense-ecosystem
- Record: https://forck.live/items/182-accelerating-the-cyber-defense-ecosystem-that-protects-us-all
- Subject: GPT / ChatGPT / API

OpenAI announces that leading security firms and enterprises are joining its Trusted Access for Cyber initiative, utilizing the GPT-5.4-Cyber model and receiving $10 million in API grants to enhance global cyber defense.

### Replit — Making scroll-stopping videos with Replit Animation

- Company: Replit (replit.com)
- Announced: 2026-04-15T22:41:04+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://replit.com/blog/viral-videos-replit-animation
- Record: https://forck.live/items/17912-making-scroll-stopping-videos-with-replit-animation
- Subject: Replit Agent

Replit published a guide on using Replit Animation, a tool launched in February that creates motion-style videos in minutes. The guide covers workflows from prompting to exporting, including iteration, canvas use, transitions, and music. Replit reports that videos made with the tool have generated over 10 million organic impressions.

### Google DeepMind — Gemini 3.1 Flash TTS: the next generation of expressive AI speech

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-04-15T16:03:19+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://deepmind.google/blog/gemini-3-1-flash-tts-the-next-generation-of-expressive-ai-speech/
- Record: https://forck.live/items/1240-gemini-3-1-flash-tts-the-next-generation-of-expressive-ai-speech
- Subject: Gemini

Google DeepMind announces its newest audio model that introduces granular audio tags for precise control over expressive AI speech generation.

### Harvey — From First Pass to Final Sign-Off: Review Tables That Keep Legal Teams in Sync

- Company: Harvey (harvey.ai)
- Announced: 2026-04-15T15:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/collaborative-review-tables
- Record: https://forck.live/items/17583-from-first-pass-to-final-sign-off-review-tables-that-keep-legal-teams-in-sync
- Subject: Harvey

Harvey updated its review tables with real-time collaboration, conditional columns, and faster setup. The improvements include an Improve Prompt feature, automatic translation of uploaded diligence request lists into columns, and cell-level reasoning with citations. Teams can now use multi-colored flags, comments, and manual input columns to triage issues and capture nuance within the table.

### Hugging Face — Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents

- Company: Hugging Face (huggingface.co)
- Announced: 2026-04-15T12:07:25+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/ibm-research/vakra-benchmark-analysis
- Record: https://forck.live/items/1519-inside-vakra-reasoning-tool-use-and-failure-modes-of-agents
- Subject: Platform

Hugging Face and IBM Research published a blog post introducing VAKRA, a tool-grounded, executable benchmark for evaluating AI agents' reasoning and tool use in enterprise environments. The post describes the benchmark's four tasks, dataset details, and analysis of failure modes observed on different tasks.

### OpenAI — The next evolution of the Agents SDK

- Company: OpenAI (openai.com)
- Announced: 2026-04-15T10:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/the-next-evolution-of-the-agents-sdk
- Record: https://forck.live/items/183-the-next-evolution-of-the-agents-sdk
- Subject: GPT / ChatGPT / API

OpenAI updated the Agents SDK to include native sandbox execution and a model-native harness, enabling developers to build secure, long-running agents.

### H Company — Meet HoloTab by HCompany. Your AI browser companion.

- Company: H Company (hcompany.ai)
- Announced: 2026-04-15T09:25:20+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://huggingface.co/blog/Hcompany/holotab
- Record: https://forck.live/items/1520-meet-holotab-by-hcompany-your-ai-browser-companion

HCompany announces HoloTab, a free Chrome extension that uses the Holo3 computer-use AI model to automate web browsing tasks, including a 'Routines' feature for recording and replaying actions.

### Alibaba — Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to All

- Company: Alibaba (alibaba.com)
- Announced: 2026-04-15T02:00:00+00:00
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://qwen.ai/blog?id=qwen3.6-35b-a3b
- Record: https://forck.live/items/4604-qwen3-6-35b-a3b-agentic-coding-power-now-open-to-all
- Subject: Qwen

Alibaba open-sourced the Qwen3.6-35B-A3B model, a sparse mixture-of-experts model with 35 billion total parameters and 3 billion active parameters, demonstrating strong agentic coding performance and multimodal capabilities, available via Qwen Studio, API, and open weights.

### H Company — HoloTab brings a computer-use agent to the browser

- Company: H Company (hcompany.ai)
- Announced: 2026-04-15
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://hcompany.ai/newsroom/meet-holotab
- Record: https://forck.live/items/18371-holotab-brings-a-computer-use-agent-to-the-browser
- Subject: Holo

H Company released HoloTab, a Chrome extension that acts as a computer-use agent in the browser, automating tasks across websites with no setup or technical skills required. The extension builds on Holo3, the company's most advanced computer-use model, and includes a Routines feature that lets users record a task once and run it anytime. HoloTab is available as a Chrome extension.

### Gradium — Gradbot: Vibe code voice agents in 50 lines of code

- Company: Gradium (gradium.ai)
- Announced: 2026-04-15
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://gradium.ai/blog/gradbot
- Record: https://forck.live/items/18322-gradbot-vibe-code-voice-agents-in-50-lines-of-code
- Subject: Gradium TTS / Phonon

Gradium released Gradbot, an open-source framework for prototyping voice agents. It provides a Rust-based multiplexing engine that coordinates speech-to-text, LLM inference, and text-to-speech pipelines, handling turn-taking, interruption, silence detection, and asynchronous tool calling out of the box. Gradbot is available via pip install and is intended for prototyping and experimentation.

### Luma — South African Agency Boundless Delivers Mazda’s First AI-Produced Commercial Using Luma Agents

- Company: Luma (lumalabs.ai)
- Announced: 2026-04-15
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lumalabs.ai/news/Boundless-Mazda-Luma
- Record: https://forck.live/items/7967-south-african-agency-boundless-delivers-mazda-s-first-ai-produced-commercial
- Subject: Ray / Dream Machine

Luma announced that South African agency Boundless produced Mazda's first AI-produced commercial using Luma Agents, Luma's AI-powered creative workflow system. The project was delivered in less than two weeks from initial brief to final approval. Boundless has also signed an agreement with Luma to integrate the company's AI technologies into its creative workflow across future client work.

### Cognition — Devin in Windsurf

- Company: Cognition (cognition.com)
- Announced: 2026-04-15
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/devin-in-windsurf
- Record: https://forck.live/items/7668-devin-in-windsurf
- Subject: Devin / SWE

Cognition announced that Devin, its cloud agent, is now integrated into Windsurf 2.0, allowing users to delegate implementation tasks to Devin with a single click and review the resulting PRs within Windsurf.

### Baidu — Introducing ERNIE-Image

- Company: Baidu (baidu.com)
- Announced: 2026-04-15
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://ernie.baidu.com/blog/posts/ernie-image/
- Record: https://forck.live/items/4642-introducing-ernie-image
- Subject: ERNIE

Baidu announced ERNIE-Image, a text-to-image generation model with 8B DiT parameters, claiming leading performance among open-weights models.

### Cursor — Canvases

- Company: Cursor (cursor.com)
- Announced: 2026-04-15
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/04-15-26
- Record: https://forck.live/items/2619-canvases
- Subject: Cursor

Cursor can now create interactive canvases with dashboards and custom interfaces using first-party components, available in Cursor 3.1.

### Replicate — How to make remarkable videos with Seedance 2.0

- Company: Replicate (replicate.com)
- Announced: 2026-04-15
- Category: model-update
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://replicate.com/blog/seedance-2
- Record: https://forck.live/items/2466-how-to-make-remarkable-videos-with-seedance-2-0
- Subject: Platform

Replicate announces Seedance 2.0, a video generation model that can take up to 9 images, 3 video clips, 3 audio files, and a text prompt to produce videos, and provides example prompts and capabilities.

### Together AI — Parcae: Doing more with fewer parameters using stable looped models

- Company: Together AI (together.ai)
- Announced: 2026-04-15
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://www.together.ai/blog/parcae
- Record: https://forck.live/items/2358-parcae-doing-more-with-fewer-parameters-using-stable-looped-models
- Subject: Inference platform

Together AI introduces Parcae, a stable looped language model architecture that achieves the quality of a Transformer with twice the parameters, using fewer parameters by increasing recurrence. A 770M parameter Parcae matches a 1.3B parameter Transformer.

### Harvey — Two Types of Legal Work. One Agentic Platform.

- Company: Harvey (harvey.ai)
- Announced: 2026-04-14T17:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/two-types-of-legal-work-one-agentic-platform
- Record: https://forck.live/items/17584-two-types-of-legal-work-one-agentic-platform
- Subject: Harvey

Harvey describes its agentic platform for legal work, which handles both repeatable tasks (e.g., contract review, compliance checks) and complex, multi-step reasoning (e.g., analyzing contract suites, drafting memos). The company states that 700,000+ tasks are executed daily and over 50 million terms are extracted weekly, and that foundation models now enable reliable agentic execution across a broader set of tasks.

### Google DeepMind — Gemini Robotics ER-1.6 enhances reasoning to help robots navigate real-world tasks.

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-04-14T15:00:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-1-6/
- Record: https://forck.live/items/1316-gemini-robotics-er-1-6-enhances-reasoning-to-help-robots-navigate-real-world
- Subject: Gemini

Gemini Robotics-ER 1.6 is an upgrade to a reasoning-first model that helps robots understand their environment.

### Supabase — We're Transferring the Stripe Sync Engine to Stripe

- Company: Supabase (supabase.com)
- Announced: 2026-04-14
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/stripe-sync-engine-transfer
- Record: https://forck.live/items/21292-we-re-transferring-the-stripe-sync-engine-to-stripe
- Subject: Supabase

Supabase has transferred the open-source Stripe Sync Engine repository from its own GitHub organization to Stripe's. The tool, which syncs Stripe billing data into a Postgres database, remains under the Apache 2.0 license and existing integrations continue to work. The latest release includes new capabilities such as coupons sync, branching support, and immediate data syncing on install.

### Kyutai — OVIE: One View Is Enough

- Company: Kyutai (kyutai.org)
- Announced: 2026-04-14
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://kyutai.org/blog/2026-04-14-ovie
- Record: https://forck.live/items/18346-ovie-one-view-is-enough
- Subject: Moshi / Pocket TTS

Kyutai introduced OVIE, a method that generates novel views from a single photograph and is trained on unpaired internet images.

### Microsoft AI — MAI-Image-2-Efficient: Flagship Quality, 41% Lower Cost

- Company: Microsoft AI (microsoft.ai)
- Announced: 2026-04-14
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://microsoft.ai/news/mai-image-2-efficient
- Record: https://forck.live/items/8429-mai-image-2-efficient-flagship-quality-41-lower-cost
- Subject: MAI

Microsoft released MAI-Image-2-Efficient, a variant of its MAI-Image-2 text-to-image model that is 22% faster, 4x more efficient, and priced nearly 41% lower at $5 per 1M text input tokens and $19.50 per 1M image output tokens. The model is available now in Microsoft Foundry and MAI Playground, with rollout across Copilot and Bing underway.

### World Labs — Streaming 3DGS worlds on the web

- Company: World Labs (worldlabs.ai)
- Announced: 2026-04-14
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.worldlabs.ai/blog/spark-2.0
- Record: https://forck.live/items/8106-streaming-3dgs-worlds-on-the-web
- Subject: Marble / Atlas

World Labs released Spark 2.0, a dynamic 3D Gaussian Splatting renderer for the web built on THREE.js and WebGL2. The update adds a Level-of-Detail system that streams and renders large 3DGS worlds on any device including desktop, iOS, Android, and VR.

### Cognition — New self-serve plans for Devin

- Company: Cognition (cognition.com)
- Announced: 2026-04-14
- Category: pricing-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cognition.com/blog/new-self-serve-plans-for-devin
- Record: https://forck.live/items/7670-new-self-serve-plans-for-devin
- Subject: Devin / SWE

Cognition announced new self-serve pricing plans for Devin, replacing the old Core and Team plans with Free, Pro, Max, Teams, and Enterprise. The company also began charging for Ask Devin, DeepWiki, and Devin Review, which were previously free. The new Teams plan lowers the entry point from $500/month to a minimum of $80/month.

### Cognition — Introducing SWE-Check: 10x Faster Bug Detection

- Company: Cognition (cognition.com)
- Announced: 2026-04-14
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://cognition.com/blog/swe-check-10x-faster
- Record: https://forck.live/items/7669-introducing-swe-check-10x-faster-bug-detection
- Subject: Devin / SWE

Cognition, in partnership with Applied Compute, introduced SWE-Check, a reinforcement learning-trained bug detection model that matches frontier performance on in-distribution evaluations and makes progress on out-of-distribution evaluations, while being faster and cheaper.

### Anthropic — Redesigning Claude Code on desktop for parallel agents

- Company: Anthropic (anthropic.com)
- Announced: 2026-04-14
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://claude.com/blog/claude-code-desktop-redesign
- Record: https://forck.live/items/7409-redesigning-claude-code-on-desktop-for-parallel-agents
- Subject: Claude

Anthropic released a redesign of the Claude Code desktop app that includes a sidebar for managing multiple parallel sessions, a drag-and-drop layout, an integrated terminal and file editor, and performance improvements. The new app is available for all Claude Code users on Pro, Max, Team, and Enterprise plans and via the Claude API.

### Cursor — CLI Debug Mode and /btw Support

- Company: Cursor (cursor.com)
- Announced: 2026-04-14
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/04-14-26
- Record: https://forck.live/items/2620-cli-debug-mode-and-btw-support
- Subject: Cursor

Cursor CLI introduces quality-of-life improvements including Debug Mode, /btw support, configuration menu, custom status bars, and other enhancements.

### OpenAI — Trusted access for the next era of cyber defense

- Company: OpenAI (openai.com)
- Announced: 2026-04-14
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://openai.com/index/scaling-trusted-access-for-cyber-defense
- Record: https://forck.live/items/184-trusted-access-for-the-next-era-of-cyber-defense
- Subject: GPT / ChatGPT / API

OpenAI introduces GPT-5.4-Cyber as part of an expansion of its Trusted Access for Cyber program, offering it to vetted defenders with enhanced safeguards.

### Google Research — Towards developing future-ready skills with generative AI

- Company: Google Research (research.google)
- Announced: 2026-04-13T18:15:59+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/towards-developing-future-ready-skills-with-generative-ai/
- Record: https://forck.live/items/1347-towards-developing-future-ready-skills-with-generative-ai
- Subject: Research

Google Research announces Vantage, a research experiment that uses generative AI to assess future-ready skills through simulated conversations, now available on Google Labs. A study with New York University found AI scoring on par with human experts.

### Google DeepMind — Gemini Robotics-ER 1.6: Powering real-world robotics tasks through enhanced embodied reasoning

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-04-13T15:52:13+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://deepmind.google/blog/gemini-robotics-er-1-6/
- Record: https://forck.live/items/1241-gemini-robotics-er-1-6-powering-real-world-robotics-tasks-through-enhanced
- Subject: Gemini

Google DeepMind announced Gemini Robotics-ER 1.6, an update to their embodied reasoning model that improves spatial reasoning and multi-view understanding for autonomous robotics.

### Lovable — How to build a safe, scalable business on Lovable

- Company: Lovable (lovable.dev)
- Announced: 2026-04-13T13:39:14+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/how-to-build-a-safe-scalable-business-on-lovable
- Record: https://forck.live/items/12091-how-to-build-a-safe-scalable-business-on-lovable
- Subject: Lovable / AI app builder

Lovable published a guide for founders on building a secure and scalable business on its platform. It covers using the built-in security scanner, asking the agent for compliance and security reviews, and preparing for traffic spikes by optimizing code and upgrading database instances.

### OpenAI — Enterprises power agentic workflows in Cloudflare Agent Cloud with OpenAI

- Company: OpenAI (openai.com)
- Announced: 2026-04-13T06:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/cloudflare-openai-agent-cloud
- Record: https://forck.live/items/185-enterprises-power-agentic-workflows-in-cloudflare-agent-cloud-with-openai
- Subject: GPT / ChatGPT / API

Cloudflare integrates OpenAI's GPT-5.4 and Codex models into its Agent Cloud platform, enabling enterprises to build and deploy AI agents.

### Cursor — Tiled Layout and Upgraded Voice Input in the Agents Window

- Company: Cursor (cursor.com)
- Announced: 2026-04-13
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/3-1
- Record: https://forck.live/items/2621-tiled-layout-and-upgraded-voice-input-in-the-agents-window
- Subject: Cursor

Cursor 3 introduces a tiled layout for the Agents Window, upgraded voice input with batch STT, branch selection in empty state, diff-to-file navigation, filters for searching files, and various improvements and bug fixes.

### Together AI — EinsteinArena: Harnessing the collective intelligence of agents in the wild to advance science

- Company: Together AI (together.ai)
- Announced: 2026-04-13
- Category: infrastructure-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/einsteinarena
- Record: https://forck.live/items/2359-einsteinarena-harnessing-the-collective-intelligence-of-agents-in-the-wild-to
- Subject: Inference platform

Together AI introduces EinsteinArena, a platform for AI agents to collaborate on open problems. Agents have already discovered new solutions to 11 open math problems, including improving the lower bound for the Kissing Number problem in dimension 11 from 593 to 604.