# forck.live — 16–22 February 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-W08, 16–22 February 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-W08 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### OpenAI — Our First Proof submissions - Company: OpenAI (openai.com) - Announced: 2026-02-20T14:30:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/first-proof-submissions - Record: https://forck.live/items/266-our-first-proof-submissions - Subject: GPT / ChatGPT / API OpenAI shares its AI model's proof attempts for the First Proof math challenge, which tests research-grade reasoning on expert-level problems. ### Vercel — Skills Night: 69,000+ ways agents are getting smarter - Company: Vercel (vercel.com) - Announced: 2026-02-20T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/skills-night-69000-ways-agents-are-getting-smarter - Record: https://forck.live/items/20745-skills-night-69-000-ways-agents-are-getting-smarter - Subject: v0 / AI SDK / AI Gateway Vercel reported that its skills.sh ecosystem has grown to over 69,000 skills and 2 million CLI installs. The company announced security partnerships with Gen, Socket, and Snyk to audit skills, and shared developer demonstrations showing skills closing the training cutoff gap and enabling agents to drive full-stack changes. ### Sarvam AI — Introducing Indus - Company: Sarvam AI (sarvam.ai) - Announced: 2026-02-20T12:00:00+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.sarvam.ai/blogs/introducing-indus/ - Record: https://forck.live/items/18389-introducing-indus - Subject: Sarvam / Bulbul / Saaras Sarvam AI opened Indus, an interface to experience its 105B sovereign model, in limited beta. The company is gradually rolling out access on limited compute capacity, with a waitlist possible initially. ### Hugging Face — Train AI models with Unsloth and Hugging Face Jobs for FREE - Company: Hugging Face (huggingface.co) - Announced: 2026-02-20 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/unsloth-jobs - Record: https://forck.live/items/1544-train-ai-models-with-unsloth-and-hugging-face-jobs-for-free - Subject: Platform Hugging Face and Unsloth announce free credits for fine-tuning models on Hugging Face Jobs, and release a Hugging Face model training skill for coding agents (Claude Code, Codex) that automates training script generation and job submission. ### Hugging Face — GGML and llama.cpp join HF to ensure the long-term progress of Local AI - Company: Hugging Face (huggingface.co) - Announced: 2026-02-20 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/ggml-joins-hf - Record: https://forck.live/items/1543-ggml-and-llama-cpp-join-hf-to-ensure-the-long-term-progress-of-local-ai - Subject: Platform GGML and llama.cpp are joining Hugging Face to ensure the long-term progress of Local AI. ### Google Antigravity — Gemini 3.1 Pro, Building with Advanced Intelligence in Google Antigravity - Company: Google Antigravity (antigravity.google) - Announced: 2026-02-19T18:19:47+00:00 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://antigravity.google/blog/gemini-3-1-pro-in-google-antigravity - Record: https://forck.live/items/4557-gemini-3-1-pro-building-with-advanced-intelligence-in-google-antigravity - Subject: Antigravity Google Antigravity announces the integration of Gemini 3.1 Pro into its platform, highlighting advanced reasoning for complex coding tasks such as refactoring, code-based animation, and dynamic visualization. The model is also rolling out via Gemini API, Google AI Studio, Vertex AI, Gemini CLI, and Android Studio. ### Google DeepMind — Gemini 3.1 Pro: A smarter model for your most complex tasks - Company: Google DeepMind (deepmind.google) - Announced: 2026-02-19T16:06:14+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://deepmind.google/blog/gemini-3-1-pro-a-smarter-model-for-your-most-complex-tasks/ - Record: https://forck.live/items/1251-gemini-3-1-pro-a-smarter-model-for-your-most-complex-tasks - Subject: Gemini Google DeepMind introduces Gemini 3.1 Pro, a model designed for complex tasks that require more than a simple answer. ### Harvey — A New Era of Collaboration for Legal and Professional Services - Company: Harvey (harvey.ai) - Announced: 2026-02-19T13:15:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/a-new-era-of-collaboration-for-legal-and-professional-services - Record: https://forck.live/items/17609-a-new-era-of-collaboration-for-legal-and-professional-services - Subject: Harvey Harvey introduced Shared Spaces, a secure AI-native collaboration environment that allows internal teams and external partners (including non-Harvey customers) to work from the same intelligence on sensitive matters such as M&A transactions and regulatory compliance. The feature was developed in partnership with design partners including Gleiss Lutz, Deutsche Telekom, PwC, IFS, KWM, and IAG. Shared Spaces includes strict permissions, access controls, and auditability to maintain security and confidentiality. ### Vercel — Video Generation with AI Gateway - Company: Vercel (vercel.com) - Announced: 2026-02-19T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/video-generation-with-ai-gateway - Record: https://forck.live/items/20746-video-generation-with-ai-gateway - Subject: v0 / AI SDK / AI Gateway Vercel's AI Gateway now supports video generation in beta, offering text-to-video, image-to-video, first-and-last-frame, and reference-to-video capabilities through AI SDK 6 and a no-code playground. Initial models include Grok Imagine from xAI, Wan from Alibaba, Kling, and Veo from Google, each with distinct strengths such as style transfer, identity preservation, native audio, and high visual fidelity. The feature is available for Pro and Enterprise plans and paid AI Gateway users. ### Lovable — The modern website is a product, not a project - Company: Lovable (lovable.dev) - Announced: 2026-02-19T10:16:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://lovable.dev/blog/the-modern-website-is-a-product - Record: https://forck.live/items/12106-the-modern-website-is-a-product-not-a-project - Subject: Lovable / AI app builder Lovable argues that companies should treat their website as a product owned and operated by the marketing team, rather than a static project. The approach emphasizes defining constraints before pages, building in the real medium, and shipping before feeling finished. Lovable provides examples of building pages like a sales FAQ and an event invitation directly within its platform. ### Lovable — Marketing velocity is the new competitive advantage - Company: Lovable (lovable.dev) - Announced: 2026-02-19T10:07:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://lovable.dev/blog/marketing-velocity-is-the-new-competitive-advantage - Record: https://forck.live/items/12107-marketing-velocity-is-the-new-competitive-advantage - Subject: Lovable / AI app builder Lovable argues that marketing advantage now depends on how quickly teams can turn new information into shipped work that customers can react to. The post identifies three characteristics of high-velocity teams: shortening the path from insight to shipping, making it cheap to be wrong, and shipping in order to learn. It suggests auditing velocity by measuring how long it takes to get a new landing page live and how many people must touch a routine campaign change before it ships. ### OpenAI — Advancing independent research on AI alignment - Company: OpenAI (openai.com) - Announced: 2026-02-19T10:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/advancing-independent-research-ai-alignment - Record: https://forck.live/items/267-advancing-independent-research-on-ai-alignment - Subject: GPT / ChatGPT / API OpenAI commits $7.5 million to The Alignment Project to fund independent AI alignment research, aiming to strengthen global efforts to address AGI safety and security risks. ### Sakana AI — Salesforce Ventures invests in Sakana AI - Company: Sakana AI (sakana.ai) - Announced: 2026-02-19T06:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://sakana.ai/salesforce-ventures - Record: https://forck.live/items/4695-salesforce-ventures-invests-in-sakana-ai - Subject: Sakana models Salesforce Ventures has invested in Sakana AI, highlighting the startup's research capabilities and understanding of Japanese enterprise needs. ### Luma — Serviceplan Group Deploys Creative AI Across Global Operations in Partnership with Luma AI - Company: Luma (lumalabs.ai) - Announced: 2026-02-19 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://lumalabs.ai/news/serviceplan-group-deploys-creative-ai-across-global-operations-in-partnership-with-luma-ai - Record: https://forck.live/items/7970-serviceplan-group-deploys-creative-ai-across-global-operations-in-partnership - Subject: Ray / Dream Machine Serviceplan Group and Luma AI announced a strategic partnership to deploy AI for creative work across the agency group’s global operations. Luma AI will serve as Serviceplan Group’s AI technology partner, embedding AI into workflows spanning strategy, creative development, content production, and delivery. The rollout positions Serviceplan Group as the first of the world’s largest agency groups to standardize AI for creative work at scale. ### Amp — The Coding Agent Is Dead - Company: Amp (ampcode.com) - Announced: 2026-02-19 - Category: deprecation - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://ampcode.com/news/the-coding-agent-is-dead - Record: https://forck.live/items/7756-the-coding-agent-is-dead - Subject: Amp Amp announced it is deprecating its editor extensions for VS Code and Cursor, effective March 5, and will focus on its CLI tool instead, citing that newer models no longer need the hand-holding of an editor agent. ### Amp — GPT‐5.3‐Codex - Company: Amp (ampcode.com) - Announced: 2026-02-19 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://ampcode.com/news/gpt-5.3-codex - Record: https://forck.live/items/7755-gpt-5-3-codex - Subject: Amp Amp announced that GPT-5.3-Codex now powers its deep mode, claiming it is better than GPT-5.2-Codex in every dimension. ### Together AI — Consistency diffusion language models: Up to 14x faster inference without sacrificing quality - Company: Together AI (together.ai) - Announced: 2026-02-19 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.together.ai/blog/consistency-diffusion-language-models - Record: https://forck.live/items/2379-consistency-diffusion-language-models-up-to-14x-faster-inference-without - Subject: Inference platform Together AI introduces consistency diffusion language models (CDLM), a method that accelerates diffusion language model inference by combining consistency-based multi-token finalization with block-wise KV caching, achieving up to 14.5x latency speedups on math and coding tasks. ### OpenAI — Introducing OpenAI for India - Company: OpenAI (openai.com) - Announced: 2026-02-18T21:00:00+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/openai-for-india - Record: https://forck.live/items/268-introducing-openai-for-india - Subject: GPT / ChatGPT / API OpenAI is expanding its AI access in India, focusing on local infrastructure, enterprise support, and workforce skill development. ### Harvey — Introducing BigLaw Bench: Global - Company: Harvey (harvey.ai) - Announced: 2026-02-18T20:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/introducing-big-law-bench-global - Record: https://forck.live/items/17610-introducing-biglaw-bench-global - Subject: Harvey Harvey introduced BigLaw Bench: Global, an extension of its legal benchmark that adds datasets for the UK, Australia, and Spain. The benchmark was built with local legal experts and covers six core task categories including drafting, long document analysis, and public research. Harvey plans to add more countries and deepen existing datasets over time. ### Hugging Face — IBM and UC Berkeley Diagnose Why Enterprise Agents Fail Using IT-Bench and MAST - Company: Hugging Face (huggingface.co) - Announced: 2026-02-18T16:15:45+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/ibm-research/itbenchandmast - Record: https://forck.live/items/1545-ibm-and-uc-berkeley-diagnose-why-enterprise-agents-fail-using-it-bench-and-mast - Subject: Platform IBM Research and UC Berkeley collaborated to study how agentic LLM systems fail in enterprise IT automation tasks. They used MAST (Multi-Agent System Failure Taxonomy) to analyze execution traces from ITBench, a benchmark for SRE, Security, and FinOps automation. They annotated 310 ITBench SRE traces across three models: Gemini-3-Flash, Kimi-K2, and GPT-OSS-120B. Key findings include that frontier models fail cleanly with few failure modes, while large open models suffer cascading failures; incorrect verification is a strong predictor of failure across models; and Kimi-K2 shows spikes in premature termination and unawareness of termination conditions. ### Google DeepMind — A new way to express yourself: Gemini can now create music - Company: Google DeepMind (deepmind.google) - Announced: 2026-02-18T16:01:38+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://deepmind.google/blog/a-new-way-to-express-yourself-gemini-can-now-create-music/ - Record: https://forck.live/items/1252-a-new-way-to-express-yourself-gemini-can-now-create-music - Subject: Gemini The Gemini app gains a new music generation capability powered by the Lyria 3 model, allowing users to create 30-second tracks from text or image prompts. ### Replit — Ship Enterprise Data Apps Faster with Replit and Databricks - Company: Replit (replit.com) - Announced: 2026-02-18T15:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/replit-and-databricks - Record: https://forck.live/items/17919-ship-enterprise-data-apps-faster-with-replit-and-databricks - Subject: Replit Agent Replit announced an integration with Databricks Apps and Databricks AppKit that lets enterprise users build applications in Replit and deploy them into Databricks, where they inherit governance, security, and observability controls. The integration provides templates, live collaboration, and AI-powered scaffolding that follows AppKit patterns. Early access is available via a sign-up page, with gated preview for existing Replit customers. ### Harvey — 5 Workflows for IP and Patent Litigation - Company: Harvey (harvey.ai) - Announced: 2026-02-18T14:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.harvey.ai/blog/5-workflows-ip-patent-litigation - Record: https://forck.live/items/17611-5-workflows-for-ip-and-patent-litigation - Subject: Harvey Harvey published a guide describing five custom workflows for IP and patent litigation teams using its Workflow Builder. The workflows cover drafting infringement claim charts, analyzing patent office actions, generating invalidity contentions from prior art, and drafting patent license agreements. The post explains how teams can embed their own standards into reusable workflows to produce consistent work at scale. ### Vercel — We Ralph Wiggumed WebStreams to make them 10x faster - Company: Vercel (vercel.com) - Announced: 2026-02-18T13:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://vercel.com/blog/we-ralph-wiggumed-webstreams-to-make-them-10x-faster - Record: https://forck.live/items/20747-we-ralph-wiggumed-webstreams-to-make-them-10x-faster - Subject: v0 / AI SDK / AI Gateway Vercel built a library called fast-webstreams that implements the WHATWG ReadableStream, WritableStream, and TransformStream APIs backed by Node.js streams internally, removing Promise and object allocation overhead for common server-side patterns. Benchmarks show pipeThrough at ~6,200 MB/s (10x faster than native WebStreams) and the React Flight byte-stream pattern at ~1,600 MB/s (14.6x faster). The work is being contributed to Node.js via an upstream PR. ### World Labs — World Labs Announces New Funding - Company: World Labs (worldlabs.ai) - Announced: 2026-02-18 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.worldlabs.ai/blog/funding-2026 - Record: https://forck.live/items/7349-world-labs-announces-new-funding - Subject: Marble / Atlas World Labs has raised $1 billion in new funding from investors including AMD, Autodesk, Emerson Collective, Fidelity Management & Research Company, NVIDIA, and Sea. The company also announced its first product, Marble, which enables users to create 3D worlds from images, video, or text. ### Cursor — CLI Improvements and Mermaid ASCII Diagrams - Company: Cursor (cursor.com) - Announced: 2026-02-18 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://cursor.com/changelog/cli-feb-18-2026 - Record: https://forck.live/items/2632-cli-improvements-and-mermaid-ascii-diagrams - Subject: Cursor Cursor adds inline ASCII diagram rendering for Mermaid code blocks, plan-to-cloud handoff, and numerous quality-of-life improvements to its CLI, including better keyboard shortcuts, unified domain allowlisting, and reliability fixes. ### Replicate — Recraft V4: image generation with design taste - Company: Replicate (replicate.com) - Announced: 2026-02-18 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://replicate.com/blog/recraft-v4 - Record: https://forck.live/items/2468-recraft-v4-image-generation-with-design-taste - Subject: Platform Replicate announces Recraft V4, a new image generation model focused on 'design taste' for intentional visual decisions. It comes in four versions: two raster (V4 and V4 Pro) and two vector (V4 SVG and V4 Pro SVG), with different resolutions, speeds, and prices. The model can generate native SVG files with editable vector paths. ### Hugging Face — One-Shot Any Web App with Gradio's gr.HTML - Company: Hugging Face (huggingface.co) - Announced: 2026-02-18 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/gradio-html-one-shot-apps - Record: https://forck.live/items/1546-one-shot-any-web-app-with-gradio-s-gr-html - Subject: Platform Gradio 6's gr.HTML component now supports custom templates, scoped CSS, and JavaScript interactivity, allowing developers to build custom web components in a single Python file. ### OpenAI — Introducing EVMbench - Company: OpenAI (openai.com) - Announced: 2026-02-18 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/introducing-evmbench - Record: https://forck.live/items/269-introducing-evmbench - Subject: GPT / ChatGPT / API OpenAI and Paradigm introduce EVMbench, a benchmark for evaluating AI agents on detecting, patching, and exploiting high-severity smart contract vulnerabilities. ### Google Research — Teaching AI to read a map - Company: Google Research (research.google) - Announced: 2026-02-17T21:37:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://research.google/blog/teaching-ai-to-read-a-map/ - Record: https://forck.live/items/1365-teaching-ai-to-read-a-map - Subject: Research Google Research introduces MapTrace, a synthetic data generation pipeline and dataset of 2 million question-answer pairs to teach multimodal large language models to trace valid paths on maps, addressing a gap in spatial reasoning. The pipeline uses Gemini 2.5 Pro and Imagen-4 models to generate diverse maps and produce pixel-level annotations. ### Lovable — How to maintain brand control when building with Lovable - Company: Lovable (lovable.dev) - Announced: 2026-02-17T20:54:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://lovable.dev/blog/how-to-maintain-brand-control-when-build-with-lovable - Record: https://forck.live/items/12108-how-to-maintain-brand-control-when-building-with-lovable - Subject: Lovable / AI app builder Lovable published a guide on maintaining brand control when marketing teams build directly using the platform. The guide describes two approaches: Cross-Project Context, which allows projects to reference and copy files and code from other projects in a workspace to maintain consistency through reference implementations; and Design Systems (available to Enterprise), which imports and codifies a company's design system so brand constraints are enforced upstream during generation rather than through post-hoc review. ### Harvey — Sonnet 4.6, Now Live in Harvey - Company: Harvey (harvey.ai) - Announced: 2026-02-17T18:00:00+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.harvey.ai/blog/sonnet-46-now-live-in-harvey - Record: https://forck.live/items/17612-sonnet-4-6-now-live-in-harvey - Subject: Harvey Harvey has made Anthropic's Claude Sonnet 4.6 available on its platform. The model scored 87.6% on Harvey's BigLaw Bench evaluation suite, with 35% of tasks receiving perfect scores. It will roll out to eligible clients in the US and EU in the coming days, with Australian availability to be announced later. ### Google DeepMind — Accelerating discovery in India through AI-powered science and education - Company: Google DeepMind (deepmind.google) - Announced: 2026-02-17T13:42:20+00:00 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://deepmind.google/blog/accelerating-discovery-in-india-through-ai-powered-science-and-education/ - Record: https://forck.live/items/1253-accelerating-discovery-in-india-through-ai-powered-science-and-education - Subject: Gemini Google DeepMind is expanding its National Partnerships for AI initiative to India, aiming to scale AI for science and education. ### Vercel — How Stably ships AI testing agents in hours, not weeks - Company: Vercel (vercel.com) - Announced: 2026-02-17T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-stably-ships-ai-testing-agents-in-hours-not-weeks - Record: https://forck.live/items/20749-how-stably-ships-ai-testing-agents-in-hours-not-weeks - Subject: v0 / AI SDK / AI Gateway Stably, a 6-person startup, uses Vercel's platform to ship AI testing agents in hours instead of weeks, eliminating infrastructure anxiety. The company won Vercel's AI Accelerator, which provided investment and enterprise credibility. Stably's AI agents run autonomous end-to-end tests by deploying on preview URLs and reading code diffs. ### Vercel — How Ramp kept 100% uptime through 100x traffic surges on Vercel - Company: Vercel (vercel.com) - Announced: 2026-02-17T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-ramp-kept-100-uptime-through-100x-traffic-surges-on-vercel - Record: https://forck.live/items/20748-how-ramp-kept-100-uptime-through-100x-traffic-surges-on-vercel - Subject: v0 / AI SDK / AI Gateway Ramp rebuilt its marketing site on Next.js and Vercel, migrating high-traffic pages incrementally with A/B testing against Core Web Vitals. The site maintained 100% uptime through 100x traffic surges during stunt-marketing events. Ramp is now using the Vercel AI SDK to build semantic search across internal assets. ### Odyssey — Why We Must Build World Models - Company: Odyssey (odyssey.systems) - Announced: 2026-02-17 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://odyssey.systems/why-we-must-build-world-models - Record: https://forck.live/items/18600-why-we-must-build-world-models - Subject: Odyssey world models From naive software to systems capable of interacting intelligently with us and our world, unlocking intelligence and scientific breakthroughs Oliver Cameron February 17th, 2026 Humans have been trying to model the world since our beginning. ### Gradium — InteractionLabs (Ongo) and Gradium Partner to Redefine Human-Robot Interaction - Company: Gradium (gradium.ai) - Announced: 2026-02-17 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://gradium.ai/blog/interactionlabs-gradium-partnership - Record: https://forck.live/items/18326-interactionlabs-ongo-and-gradium-partner-to-redefine-human-robot-interaction - Subject: Gradium TTS / Phonon Gradium and InteractionLabs announced a partnership to integrate Gradium's speech-to-text and text-to-speech models into InteractionLabs' robot Ongo, aiming to make human-robot interaction more natural and emotionally expressive. ### TII — Falcon‑H1R-FP8: Accelerating Inference with Quantized Precision - Company: TII (tii.ae) - Announced: 2026-02-16T08:00:00+00:00 - Category: model-update - Coverage: not counted - Announcement: yes - Group: models - Source: https://falcon-lm.github.io/blog/falcon-h1r-7b-fp8/ - Record: https://forck.live/items/18411-falcon-h1r-fp8-accelerating-inference-with-quantized-precision - Subject: Falcon LLM TII released Falcon H1R 7B FP8, a fully quantized version of the Falcon H1R 7B model with both weights and activations in FP8 format. The model preserves BF16-level accuracy (AIME25 drops 0.8%, LCB-v6 falls 1%, GPQA-D shows 0.1% difference) while achieving 1.2×–1.5× throughput improvement and reducing GPU memory footprint by about 44% (from 14.2 GB to 7.9 GB). The quantization was performed using NVIDIA Model Optimizer with per-tensor FP8 post-training quantization and also quantizes the KV cache to FP8.