# forck.live — 12–18 January 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-W03, 12–18 January 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-W03
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Lovable — From Buy to Build: How eXp Realty cancelled millions in SaaS contracts for custom software built with Lovable

- Company: Lovable (lovable.dev)
- Announced: 2026-01-18T22:01:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://lovable.dev/blog/exprealty
- Record: https://forck.live/items/12111-from-buy-to-build-how-exp-realty-cancelled-millions-in-saas-contracts-for
- Subject: Lovable / AI app builder

eXp Realty, a large real estate brokerage with over 83,000 agents, used Lovable to build custom software that replaced expensive SaaS tools. They created a custom property search website platform for 26 countries, cutting support tickets by 85%, and built an internal communications platform called 'The Hub' that eliminated approximately $1 million per year in per-seat licensing. They also replaced ChatGPT Enterprise custom chatbots with OpenAI API workflows, saving another approximately $1 million per year.

### Sakana AI — RePo: Language Models with Context Re-Positioning

- Company: Sakana AI (sakana.ai)
- Announced: 2026-01-18T15:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://sakana.ai/repo
- Record: https://forck.live/items/4699-repo-language-models-with-context-re-positioning
- Subject: Sakana models

Sakana AI announces RePo, a method that allows language models to dynamically reorganize their context by learning to assign positions based on content relevance, improving robustness on noisy contexts, structured data, and long-range dependencies.

### OpenAI — A business that scales with the value of intelligence

- Company: OpenAI (openai.com)
- Announced: 2026-01-18T10:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/a-business-that-scales-with-the-value-of-intelligence
- Record: https://forck.live/items/308-a-business-that-scales-with-the-value-of-intelligence
- Subject: GPT / ChatGPT / API

OpenAI's business model is described as scaling with intelligence, covering subscriptions, API, ads, commerce, and compute, driven by ChatGPT adoption.

### Harvey — Rebuilding Harvey's Design System From the Ground Up

- Company: Harvey (harvey.ai)
- Announced: 2026-01-16T14:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/rebuilding-harveys-design-system-from-the-ground-up
- Record: https://forck.live/items/17624-rebuilding-harvey-s-design-system-from-the-ground-up
- Subject: Harvey

Harvey Engineering, Product, and Design teams rebuilt their design system to address color and component sprawl. They established a semantic token architecture, automated token maintenance via a GitHub Action syncing Figma variables to code, and recalibrated the color palette for WCAG contrast in light and dark modes. The new system aligns naming conventions between design and code and is intended to scale without accumulating technical debt.

### Google DeepMind — D4RT: Teaching AI to see the world in four dimensions

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-01-16T10:39:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://deepmind.google/blog/d4rt-teaching-ai-to-see-the-world-in-four-dimensions/
- Record: https://forck.live/items/1257-d4rt-teaching-ai-to-see-the-world-in-four-dimensions
- Subject: Gemini

Google DeepMind announces D4RT, a method for unified, efficient 4D reconstruction and tracking that is up to 300x faster than prior methods.

### Cursor — CLI Agent Modes and Cloud Handoff

- Company: Cursor (cursor.com)
- Announced: 2026-01-16
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://cursor.com/changelog/cli-jan-16-2026
- Record: https://forck.live/items/2636-cli-agent-modes-and-cloud-handoff
- Subject: Cursor

Cursor CLI adds Plan mode and Ask mode for interactive planning and code exploration, introduces Handoff to Cloud Agents to continue conversations on web or mobile, provides word-level inline diffs, and enables one-click MCP authentication. The release also includes improvements like hooks, usage stats, web tools, better terminal support, and several bug fixes.

### OpenAI — Introducing ChatGPT Go, now available worldwide

- Company: OpenAI (openai.com)
- Announced: 2026-01-16
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/introducing-chatgpt-go
- Record: https://forck.live/items/310-introducing-chatgpt-go-now-available-worldwide
- Subject: GPT / ChatGPT / API

ChatGPT Go is now available worldwide, expanding access to GPT-5.2 Instant with higher usage limits and longer memory.

### OpenAI — Our approach to advertising and expanding access to ChatGPT

- Company: OpenAI (openai.com)
- Announced: 2026-01-16
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/our-approach-to-advertising-and-expanding-access
- Record: https://forck.live/items/309-our-approach-to-advertising-and-expanding-access-to-chatgpt
- Subject: GPT / ChatGPT / API

OpenAI plans to test advertising in the U.S. for ChatGPT's free and Go tiers to expand affordable access to AI worldwide, while protecting privacy, trust, and answer quality.

### Google Research — Unlocking health insights: Estimating advanced walking metrics with smartwatches

- Company: Google Research (research.google)
- Announced: 2026-01-15T22:56:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/unlocking-health-insights-estimating-advanced-walking-metrics-with-smartwatches/
- Record: https://forck.live/items/1376-unlocking-health-insights-estimating-advanced-walking-metrics-with-smartwatches
- Subject: Research

A deep learning model for estimating spatio-temporal gait metrics (e.g., walking speed, step length) from smartwatch IMU data. The model uses a temporal convolutional network with multi-head output and was validated on 246 participants with ~70,000 walking segments, showing strong correlation (Pearson r>0.80) and excellent reliability (ICC>0.80) for most metrics compared to a lab-grade system. This is primarily a research paper sharing findings.

### Replit — Mobile Apps on Replit: Idea to App Store in Minutes

- Company: Replit (replit.com)
- Announced: 2026-01-15T20:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/mobile-apps
- Record: https://forck.live/items/17921-mobile-apps-on-replit-idea-to-app-store-in-minutes
- Subject: Replit Agent

Replit launched Mobile Apps on Replit, a new capability that lets users build and publish mobile apps using natural language descriptions. The platform generates full-stack, production-ready apps with AI capabilities, built-in databases, and connectors, and allows publishing to the App Store with three clicks. All Replit users can now build mobile apps from both the Replit desktop and mobile app.

### Vercel — AWS databases are now live on the Vercel Marketplace and v0

- Company: Vercel (vercel.com)
- Announced: 2026-01-15T13:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/aws-databases-are-now-live-on-the-vercel-marketplace-and-v0
- Record: https://forck.live/items/20763-aws-databases-are-now-live-on-the-vercel-marketplace-and-v0
- Subject: v0 / AI SDK / AI Gateway

Vercel made Aurora PostgreSQL, Amazon DynamoDB, and Aurora DSQL available to provision directly from the Vercel Marketplace and through v0. Developers can set up a database with a guided flow and automatic project configuration, and v0 can create and connect AWS accounts and databases during app building. New AWS teams receive a free starter plan and $100 in credits.

### ElevenLabs — MasterClass brings AI instructors to life with ElevenLabs

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-01-15T12:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://elevenlabs.io/blog/masterclass
- Record: https://forck.live/items/3374-masterclass-brings-ai-instructors-to-life-with-elevenlabs
- Subject: Voice AI

MasterClass is using ElevenLabs to create AI instructors.

### Harvey — Harvey Hires VP of EMEA Sales and Announces New Paris Office

- Company: Harvey (harvey.ai)
- Announced: 2026-01-15T08:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/harvey-hires-vp-of-emea-sales-and-announces-new-paris-office
- Record: https://forck.live/items/17625-harvey-hires-vp-of-emea-sales-and-announces-new-paris-office
- Subject: Harvey

Harvey announced the hiring of Jorge Bestard as Vice President of EMEA Sales and plans to open a new office in Paris in Q2, continuing its investment in the European market. The company has grown its London team to over 75 employees and expanded into Spain and Germany, with strategic investors EQT and Evantic. Harvey already partners with French law firms including Adecco, August Debouzy, Berenice, Bredin Prat, and Vinci Energies.

### OpenAI — Investing in Merge Labs

- Company: OpenAI (openai.com)
- Announced: 2026-01-15T07:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/investing-in-merge-labs
- Record: https://forck.live/items/311-investing-in-merge-labs
- Subject: GPT / ChatGPT / API

OpenAI is investing in Merge Labs to support new brain computer interfaces that bridge biological and artificial intelligence.

### Replit — How Replit Secures AI-Generated Code [white paper]

- Company: Replit (replit.com)
- Announced: 2026-01-15T01:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/securing-ai-generated-code
- Record: https://forck.live/items/17923-how-replit-secures-ai-generated-code-white-paper
- Subject: Replit Agent

Replit's white paper presents findings from controlled experiments comparing AI-only security scans with hybrid approaches that combine deterministic static analysis, dependency scanning, and LLM-based reasoning. The research finds that AI-only scans are nondeterministic, prompt-sensitive, and miss dependency vulnerabilities, while static analysis provides consistency and a hybrid architecture is essential for comprehensive protection.

### OpenAI — Building with the GPT-5 Model Series: A Practical Guide

- Company: OpenAI (openai.com)
- Announced: 2026-01-15T00:09:56+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/business/guides-and-resources/a-practical-guide-to-building-with-ai
- Record: https://forck.live/items/18797-building-with-the-gpt-5-model-series-a-practical-guide
- Subject: GPT / ChatGPT / API

OpenAI published a practical guide for startups on building with GPT-5, covering migration to the Responses API, prompt optimization, and new steering controls such as reasoning_effort and verbosity. The guide states GPT-5 scored 74.9% on SWE-bench Verified and is described as faster, smarter, and more steerable than previous models.

### Supabase — When to use Read Replicas vs. bigger compute

- Company: Supabase (supabase.com)
- Announced: 2026-01-15
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://supabase.com/blog/read-replicas-vs-bigger-compute
- Record: https://forck.live/items/21308-when-to-use-read-replicas-vs-bigger-compute
- Subject: Supabase

Supabase's guide explains how to diagnose database slowdowns and decide between vertical scaling (bigger compute) and horizontal read scaling (Read Replicas), covering query patterns, read/write ratios, and CPU/memory metrics.

### Gradium — Powering Wonderful's Voice Agents

- Company: Gradium (gradium.ai)
- Announced: 2026-01-15
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://gradium.ai/blog/wonderful
- Record: https://forck.live/items/18332-powering-wonderful-s-voice-agents
- Subject: Gradium TTS / Phonon

Gradium announced it is powering real-time voice agents on Wonderful's platform, citing shared values of local implantation, global reach, and execution focus.

### Amp — Tab, Tab, Dead

- Company: Amp (ampcode.com)
- Announced: 2026-01-15
- Category: deprecation
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/tab-tab-dead
- Record: https://forck.live/items/7761-tab-tab-dead
- Subject: Amp

Amp is removing Amp Tab, its code completion feature, stating that the era of tab completion is ending as agents now write most code. Amp Tab will continue to work until the end of January 2026, after which users are directed to Cursor, Copilot, or Zed for inline completions.

### Baidu — ERNIE-5.0 Tops LMArena Text Leaderboard as No.1 Chinese Model!

- Company: Baidu (baidu.com)
- Announced: 2026-01-15
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://ernie.baidu.com/blog/posts/ernie-5.0-0110-release-on-lmarena/
- Record: https://forck.live/items/4645-ernie-5-0-tops-lmarena-text-leaderboard-as-no-1-chinese-model
- Subject: ERNIE

ERNIE-5.0-0110 ranked first among Chinese models and eighth globally on the LMArena Text leaderboard, and second globally on Math. The model is no longer labeled as Preview.

### Hugging Face — Open Responses: What you need to know

- Company: Hugging Face (huggingface.co)
- Announced: 2026-01-15
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/open-responses
- Record: https://forck.live/items/1564-open-responses-what-you-need-to-know
- Subject: Platform

Hugging Face announces Open Responses, an open inference standard based on OpenAI's Responses API, designed for agentic workloads and structured streaming.

### OpenAI — Strengthening the U.S. AI supply chain through domestic manufacturing

- Company: OpenAI (openai.com)
- Announced: 2026-01-15
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/strengthening-the-us-ai-supply-chain
- Record: https://forck.live/items/312-strengthening-the-u-s-ai-supply-chain-through-domestic-manufacturing
- Subject: GPT / ChatGPT / API

OpenAI launches a new RFP to strengthen the U.S. AI supply chain by accelerating domestic manufacturing, creating jobs, and scaling AI infrastructure.

### OpenAI — How OpenAI Uses Codex to Accelerate Engineering

- Company: OpenAI (openai.com)
- Announced: 2026-01-14T21:27:09+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/business/guides-and-resources/how-openai-uses-codex
- Record: https://forck.live/items/18799-how-openai-uses-codex-to-accelerate-engineering
- Subject: GPT / ChatGPT / API

OpenAI published a guide detailing how its own engineering teams use Codex for code understanding, refactoring, performance optimization, test coverage, and development velocity, based on interviews and internal usage data.

### Harvey — Harvey Expands Law School Program to Sydney

- Company: Harvey (harvey.ai)
- Announced: 2026-01-14T21:00:00+00:00
- Category: availability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/harvey-expands-law-school-program-to-sydney
- Record: https://forck.live/items/17626-harvey-expands-law-school-program-to-sydney
- Subject: Harvey

Harvey announced the expansion of its Law Schools Program to Australia, partnering with the University of Sydney Law School and the University of Technology Sydney (UTS) Faculty of Law. The program provides students and faculty access to Harvey's platform starting in February, as part of Harvey's growing investment in the Asia-Pacific region.

### OpenAI — OpenAI partners with Cerebras

- Company: OpenAI (openai.com)
- Announced: 2026-01-14T14:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/cerebras-partnership
- Record: https://forck.live/items/313-openai-partners-with-cerebras
- Subject: GPT / ChatGPT / API

OpenAI announces a partnership with Cerebras to add 750MW of high-speed AI compute, aiming to reduce inference latency and improve ChatGPT speed for real-time AI workloads.

### Vercel — Introducing: React Best Practices

- Company: Vercel (vercel.com)
- Announced: 2026-01-14T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/introducing-react-best-practices
- Record: https://forck.live/items/20765-introducing-react-best-practices
- Subject: v0 / AI SDK / AI Gateway

Vercel released a structured repository of React and Next.js performance best practices, containing 40+ rules across 8 categories ordered by impact, designed for use by AI agents and human developers.

### Vercel — Use Perplexity Web Search with Vercel AI Gateway

- Company: Vercel (vercel.com)
- Announced: 2026-01-14T13:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/use-perplexity-web-search-with-vercel-ai-gateway
- Record: https://forck.live/items/20764-use-perplexity-web-search-with-vercel-ai-gateway
- Subject: v0 / AI SDK / AI Gateway

Vercel added provider-agnostic web search to AI Gateway, allowing any supported model to search the web in real-time via a single line of code using Perplexity. The feature works with OpenAI, Anthropic, Google, and other providers, and is priced at $5 per 1,000 requests.

### Harvey — Harvey In Practice: How M&A Teams Use Harvey Across the Deal Lifecycle

- Company: Harvey (harvey.ai)
- Announced: 2026-01-14T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.harvey.ai/blog/harvey-in-practice-how-m-and-a-teams-use-harvey
- Record: https://forck.live/items/17627-harvey-in-practice-how-m-a-teams-use-harvey-across-the-deal-lifecycle
- Subject: Harvey

A look at how transactional teams use Harvey throughout the M&A lifecycle, from early-stage diligence to signing and closing. M&A transactions move fast.

### ElevenLabs — Deutsche Telekom and ElevenLabs announce partnership

- Company: ElevenLabs (elevenlabs.io)
- Announced: 2026-01-14T12:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://elevenlabs.io/blog/deutsche-telekom-and-elevenlabs-announce-partnership
- Record: https://forck.live/items/3367-deutsche-telekom-and-elevenlabs-announce-partnership
- Subject: Voice AI

Deutsche Telekom and ElevenLabs announce a partnership.

### Amp — Painter

- Company: Amp (ampcode.com)
- Announced: 2026-01-14
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/painter
- Record: https://forck.live/items/7762-painter
- Subject: Amp

Amp added the ability to generate and edit images using a new painter tool, which is powered by Gemini 3 Pro Image (also called Nano Banana Pro).

### Z.ai — GLM-Image

- Company: Z.ai (z.ai)
- Announced: 2026-01-14
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://docs.z.ai/release-notes/new-released#2026-01-14
- Record: https://forck.live/items/4454-glm-image
- Subject: GLM

Z.ai announced the launch of GLM-Image, a new image generation model built on a multimodal architecture and trained on domestic chips, combining autoregressive semantic understanding with diffusion-based decoding for high-quality controllable visual generation, with enhanced performance in knowledge-intensive scenarios and stable text rendering.

### Google Research — Hard-braking events as indicators of road segment crash risk

- Company: Google Research (research.google)
- Announced: 2026-01-13T22:44:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/hard-braking-events-as-indicators-of-road-segment-crash-risk/
- Record: https://forck.live/items/1377-hard-braking-events-as-indicators-of-road-segment-crash-risk
- Subject: Research

Google Research establishes a positive association between hard-braking events (HBEs) collected via Android Auto and actual road segment crash rates, suggesting HBEs can be used as leading measures for road safety assessment.

### Google Research — Next generation medical image interpretation with MedGemma 1.5 and medical speech to text with MedASR

- Company: Google Research (research.google)
- Announced: 2026-01-13T20:57:16+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://research.google/blog/next-generation-medical-image-interpretation-with-medgemma-15-and-medical-speech-to-text-with-medasr/
- Record: https://forck.live/items/1378-next-generation-medical-image-interpretation-with-medgemma-1-5-and-medical
- Subject: Research

Google Research announces MedGemma 1.5 4B, an update to the open MedGemma model with improved support for high-dimensional medical imaging (CT, MRI, histopathology), longitudinal imaging, anatomical localization, and medical document understanding. They also describe MedASR, a new open medical speech-to-text model. Both remain free for research and commercial use.

### Google Research — Dynamic surface codes open new avenues for quantum error correction

- Company: Google Research (research.google)
- Announced: 2026-01-13T17:32:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://research.google/blog/dynamic-surface-codes-open-new-avenues-for-quantum-error-correction/
- Record: https://forck.live/items/1379-dynamic-surface-codes-open-new-avenues-for-quantum-error-correction
- Subject: Research

Google Quantum AI announces the experimental demonstration of dynamic surface codes for quantum error correction, presenting three new circuit types (hexagonal, walking, iSWAP) that reduce couplers, limit leakage, and avoid dropouts, as published in Nature Physics.

### Replit — ChatPRD + Replit: From Spec to Running App in Minutes

- Company: Replit (replit.com)
- Announced: 2026-01-13T17:01:51+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/chatprd
- Record: https://forck.live/items/17924-chatprd-replit-from-spec-to-running-app-in-minutes
- Subject: Replit Agent

Replit announced an integration with ChatPRD that allows product managers to convert a product requirements document into a deployed full-stack application using Replit Agent. The workflow transfers the PRD automatically to Replit, where the agent builds and deploys a working prototype without manual setup. The integration aims to reduce the time from spec to working app from weeks to minutes.

### Google DeepMind — Veo 3.1 Ingredients to Video: More consistency, creativity and control

- Company: Google DeepMind (deepmind.google)
- Announced: 2026-01-13T17:00:18+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://deepmind.google/blog/veo-3-1-ingredients-to-video-more-consistency-creativity-and-control/
- Record: https://forck.live/items/1258-veo-3-1-ingredients-to-video-more-consistency-creativity-and-control
- Subject: Gemini

Veo update generates lively, dynamic clips and supports vertical video generation.

### OpenAI — Zenken boosts a lean sales team with ChatGPT Enterprise

- Company: OpenAI (openai.com)
- Announced: 2026-01-13T16:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://openai.com/index/zenken
- Record: https://forck.live/items/314-zenken-boosts-a-lean-sales-team-with-chatgpt-enterprise
- Subject: GPT / ChatGPT / API

OpenAI published a customer story about Zenken using ChatGPT Enterprise to improve sales performance, reduce preparation time, and increase proposal success rates.

### Vercel — Nick Bogaty joins Vercel as Chief Revenue Officer

- Company: Vercel (vercel.com)
- Announced: 2026-01-13T13:00:00+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://vercel.com/blog/nick-bogaty-joins-vercel-as-chief-revenue-officer
- Record: https://forck.live/items/20766-nick-bogaty-joins-vercel-as-chief-revenue-officer
- Subject: v0 / AI SDK / AI Gateway

Vercel announced the appointment of Nick Bogaty as Chief Revenue Officer. Bogaty, who joins from Adobe, will lead the global Sales team as the company focuses on building an AI-forward go-to-market organization.

### Kyutai — Pocket TTS: a high-quality TTS with voice cloning that runs on CPU

- Company: Kyutai (kyutai.org)
- Announced: 2026-01-13
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://kyutai.org/blog/2026-01-13-pocket-tts
- Record: https://forck.live/items/18349-pocket-tts-a-high-quality-tts-with-voice-cloning-that-runs-on-cpu
- Subject: Moshi / Pocket TTS

A tiny TTS with any voice you like.

### Amp — Stick a Fork in It, It's Done

- Company: Amp (ampcode.com)
- Announced: 2026-01-13
- Category: deprecation
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/stick-a-fork-in-it
- Record: https://forck.live/items/7764-stick-a-fork-in-it-it-s-done
- Subject: Amp

Amp is removing the Fork command from its product. The company recommends using handoff and thread mentions as alternatives.

### Amp — Handoff, Please

- Company: Amp (ampcode.com)
- Announced: 2026-01-13
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/ask-to-handoff
- Record: https://forck.live/items/7763-handoff-please
- Subject: Amp

Amp now allows users to request a handoff to a new thread by asking the agent directly, instead of requiring a button click or command. The agent starts a new thread with relevant context and continues working.

### Together AI — Learn how Cursor partnered with Together AI to deliver real-time, low-latency inference at scale

- Company: Together AI (together.ai)
- Announced: 2026-01-13
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.together.ai/blog/learn-how-cursor-partnered-with-together-ai-to-deliver-real-time-low-latency-inference-at-scale
- Record: https://forck.live/items/2388-learn-how-cursor-partnered-with-together-ai-to-deliver-real-time-low-latency
- Subject: Inference platform

Cursor and Together AI partner to deliver real-time, low-latency inference for Cursor's AI coding platform using NVIDIA Blackwell GPUs, custom kernels, and a quantization pipeline to enable fast model iteration.

### Google Research — NeuralGCM harnesses AI to better simulate long-range global precipitation

- Company: Google Research (research.google)
- Announced: 2026-01-12T17:52:00+00:00
- Category: model-update
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://research.google/blog/neuralgcm-harnesses-ai-to-better-simulate-long-range-global-precipitation/
- Record: https://forck.live/items/1380-neuralgcm-harnesses-ai-to-better-simulate-long-range-global-precipitation
- Subject: Research

Google Research describes an update to NeuralGCM, a hybrid atmospheric model that combines physics-based modeling with a neural network. The new version is trained directly on NASA satellite-based precipitation observations (2001–2018) instead of reanalysis data, leading to improved simulations of average precipitation, precipitation extremes (especially the top 0.1% of rainfall), and the daily weather cycle. At 280 km resolution, it outperforms a leading operational model from ECMWF on 2–15 day forecasts and shows improvements for multi-decadal climate simulations.

### Harvey — The Brief: January 2026

- Company: Harvey (harvey.ai)
- Announced: 2026-01-12T16:00:00+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.harvey.ai/blog/the-brief-january-2026
- Record: https://forck.live/items/17628-the-brief-january-2026
- Subject: Harvey

Harvey made GPT-5.2 available across Assistant, Vault, and Workflow Builder, improved document understanding by reading tables in PDFs and Word files, and added over 50 legal research sources covering jurisdictions including the US, UK, and multiple European countries.

### Vercel — How Mux shipped durable video workflows with their @mux/ai SDK

- Company: Vercel (vercel.com)
- Announced: 2026-01-12T13:00:00+00:00
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://vercel.com/blog/how-mux-shipped-durable-video-workflows-with-their-mux-ai-sdk
- Record: https://forck.live/items/20767-how-mux-shipped-durable-video-workflows-with-their-mux-ai-sdk
- Subject: v0 / AI SDK / AI Gateway

Mux built @mux/ai, an open-source SDK for AI features on video infrastructure, using Vercel's Workflow DevKit to provide durable execution. The SDK uses "use workflow" and "use step" directives that enable automatic retries, state persistence, and observability when deployed in a Workflow DevKit environment, while remaining a normal Node package otherwise. Workflows can be developed locally and deployed to Vercel with zero infrastructure configuration and built-in observability.

### Inception — SearchBlox + Inception: Real-Time GenAI Search at Enterprise Scale

- Company: Inception (inceptionlabs.ai)
- Announced: 2026-01-12
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.inceptionlabs.ai/blog/searchblox-and-inception
- Record: https://forck.live/items/18525-searchblox-inception-real-time-genai-search-at-enterprise-scale
- Subject: Mercury

SearchBlox has integrated Inception's Mercury diffusion LLM into its SearchAI platform, enabling sub-second GenAI responses for enterprise workloads. The integration requires no pipeline refactoring and promises 60 to 90% lower inference costs compared to autoregressive LLMs. Customers can deploy Mercury on AWS Bedrock or Azure Foundry for secure private-cloud use.

### Genspark — Seeing AGI (6): Future Education Starts at Home — A Letter to All Parents

- Company: Genspark (genspark.ai)
- Announced: 2026-01-12
- Category: not stated
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.genspark.ai/blog/seeing-agi-impact6
- Record: https://forck.live/items/18262-seeing-agi-6-future-education-starts-at-home-a-letter-to-all-parents
- Subject: Genspark

### Amp — TODOs Are Done

- Company: Amp (ampcode.com)
- Announced: 2026-01-12
- Category: deprecation
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://ampcode.com/news/todos-are-done
- Record: https://forck.live/items/7765-todos-are-done
- Subject: Amp

Amp removed the TODOs feature from its product, stating that with Opus 4.5 the feature is no longer needed and its removal improves speed.

### Together AI — Inside multi-node training: How to scale model training across GPU clusters

- Company: Together AI (together.ai)
- Announced: 2026-01-12
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.together.ai/blog/multi-node-gpu-training
- Record: https://forck.live/items/2389-inside-multi-node-training-how-to-scale-model-training-across-gpu-clusters
- Subject: Inference platform

This article is a technical guide explaining multi-node GPU training, covering parallelism strategies, network interconnects, checkpointing, and practical steps for scaling model training across GPU clusters, with a production example using Qwen2.5-72B.