Search the archive

Register now

TRIANGLE SPLATS FROM VIDEO DIFFUSION LATENTS

Google's FLAT introduces a feedforward method that decodes triangle splats directly from video diffusion latents, improving geometric accuracy over 3D Gaussian-based approaches.

QWEN-AGENTWORLD

Alibaba introduced Qwen-AgentWorld, a family of language world models trained on more than 10 million environment interaction trajectories to simulate agentic environments across multiple domains.

ORCA

Functions as an open-source Agent Development Environment designed to manage and orchestrate fleets of parallel coding agents simultaneously.

BONUS, FULLY REMOTE)

We're hiring a senior PMM to own product marketing at TLDR. You'll define our positioning, build out sales enablement, and lead every launch. Learn more [20].

AS AI COMPANIES RACE FOR POWER, AMAZON AND GOOGLE HAVE THE LEAD

Amazon has an incumbent advantage in the race for hyperscalers to get their hands on more electricity. It has been building tons of data centers over the past two decades. The company is expected to add the most data center and power capacity in the US through 2030. However, Google will have significantly closed its gap with Amazon by that time.

ANTHROPIC AND ALIBABA LAUNCH JOINT AI MODEL DISTILLATION CAMPAIGN

Anthropic and Alibaba have initiated a collaborative open-source framework focused on distilling advanced reasoning intelligence from frontier models into hyper-efficient edge models. The partnership leverages Anthropic's safety-alignment techniques alongside Alibaba's massive cloud infrastructure to compress compute footprints without severe capability degradation.

for AI in this virtual panel.

FABLE 5 HAS NOW REPORTEDLY ALSO REAPPEARED IN AMAZON BEDROCK

Claude Code v2.2.190 introduces several string changes that hint at preparations for the return of Fable 5.

ANTHROPIC VETERANS' STARTUP SEEKS TO HELP SCIENTISTS DEVELOP THEIR OWN AI

Mirendil has raised $200 million in seed funding to make and distribute AI that accelerates AI research for everyone.

OPENAI UPDATES GPT-5.5 INSTANT TO MAKE CHATGPT MORE NATURAL AND USEFUL

OpenAI has initiated the rollout of an upgraded GPT-5.5 Instant model directly inside ChatGPT for both paid and free tiers.

PERPLEXITY COMPUTER FOR COUNSEL

Perplexity launched Computer for Counsel, an AI-driven legal operations tool designed to automate administrative research, document gathering, and contract triage.

CLAUDE TAG

Anthropic introduced Claude Tag, a Slack-based workflow that lets teams assign tasks to Claude, connect it to tools and codebases, and have it retain context across channels. The company said the system had become a core part of internal operations, with its product team using it to generate much of their code and assist with analytics, support, and debugging tasks.

BYTEDANCE'S NEW AI VIDEO MODEL CAN MAKE 30-SECOND CLIPS FROM A SINGLE PROMPT

ByteDance's new Seedance 2.5 AI video generation model can generate 30-second, 4K videos with a single prompt. Users are able to provide up to 50 images, videos, or audio clips as reference pieces. Increasing the number of references gives users more control over the video creation process. The model will be available in China next month. ByteDance has not announced a release time window for other countries.

MISTRAL OCR 4: SOTA OCR FOR DOCUMENT INTELLIGENCE

INSIGHTS ON INDIRECT PROMPT INJECTION

This deep dive explores the growing focus on jailbreaks and indirect prompt injection attacks, featuring insights from Gray Swan's founders and their research. It also covered the company's role in evaluating advanced AI systems and developing security benchmarks.

BUILD REAL AGENTIC APPS USING CUGA: TWO DOZEN WORKING EXAMPLES ON A LIGHTWEIGHT HARNESS

CUGA, IBM's open-source agent harness, simplifies developing agentic apps by managing the complexities of planning, execution, and state management, allowing developers to focus on tool selection and prompts. CUGA's efficient system maintains state and corrects errors, outperforming others in benchmarks like AppWorld. Its unique features include configurable reasoning modes and integrated policy systems, enabling quick deployment from development to production while maintaining governance and flexibility.

HOW BUSINESSES ARE BUILDING SPECIALIZED AI THEY CAN TRUST

NVIDIA's Agent Toolkit empowers businesses to build specialized, customizable AI agents using open models, tools, skills, and secure runtime. These agents accelerate complex workflows across industries like life sciences, healthcare, cybersecurity, and industrial operations by integrating with existing tools and data. Companies like Cadence, Synopsys, and CrowdStrike are leveraging this technology to enhance efficiency and accuracy in specific domains.

Try Lightfield free → lightfield.app

PROMPT INJECTION AS ROLE CONFUSION

Modern large language models use role tags as both a security architecture and cognitive scaffolding. Prompt injections are driven by a flaw in how AI models perceive roles. For LLMs, everything arrives through the same channel as one long token soup, so they can't distinguish between their own thoughts and speech. Unless AI models achieve genuine role perception, injection defense will remain a perpetual whack-a-mole game.

GRAPHSIGNAL

Graphsignal is a production-scale inference profiling platform that provides essential visibility across the inference stack. It helps engineers optimize AI performance across models, engines, GPUs, and other accelerators. Graphsignal can be used with coding agents for analysis. The profiler has minimal impact on production performance, and content data is not recorded.

KREA 2 TECHNICAL REPORT

Krea 2 introduces expansive, expressive image generation models designed for creative exploration, overcoming limitations of default aesthetics. It employs a multi-stage training process with advanced architectures and extensive data curation to enhance stylistic diversity and user control. Key innovations include a prompt expander and style-reference system, allowing nuanced text and image inputs to generate diverse visual outputs.

UNLIMITED OCR WORKS

Unlimited OCR is a model designed to emulate human parsing working memory. It uses DeepSeek OCR as a baseline and combines it with a constant KV cache design. Unlimited OCR can transcribe dozens of pages of documents in a single forward pass under a standard maximum length of 32K. The technique used to develop Unlimited OCR is equally applicable to tasks such as ASR and translation.

BONUS, FULLY REMOTE)

We're hiring a senior PMM to own product marketing at TLDR. You'll define our positioning, build out sales enablement, and lead every launch. Learn more [19].

US PRESSES META TO AGREE TO AI REVIEWS AS SECURITY CONCERNS RISE

The Trump administration is pressing Meta to submit its AI models for voluntary review. Meta is the only major AI developer in the US that has not reached an agreement to voluntarily share its models with the federal government for review. The review involved evaluating models' abilities and vulnerabilities. Meta's policy team has been negotiating with the Commerce Department about how to proceed, but it is unclear whether they will be able to reach an agreement.

OPENAI PREPARES BIDIRECTIONAL VOICE MODE FOR ROLLOUT ON CHATGPT

OpenAI has started rolling out Bidirectional Voice Mode for ChatGPT. The company's new audio generation model, Bidi 1, lets the assistant speak, hear, and listen at the same time. It is able to hold the thread of a whole conversation and switch tasks on the fly if interrupted. The model can sing and beatbox, but there are some tight copyright restrictions. OpenAI has yet to make a formal announcement about the model, but some users are already seeing it in their model selectors.

INTRODUCING ENGRAM: SCALING COMPUTE ON YOUR CONTEXT

Engram is building AI models that continuously learn from a user's private context, such as documents, chats, code, and knowledge bases, instead of repeatedly rereading the same information in every session.

A NEW ERA OF SOFTWARE QUALITY STARTS TODAY

Momentic's new platform update offers autonomous QA testing, enabling teams to define product behavior and adapt tests to changes automatically.

FLUREE DB

Fluree DB is a graph database for data with integrated vector, text, and geo search.

NVIDIA AND AWS COLLABORATE TO BRING AI TO PRODUCTION AT SCALE

NVIDIA and AWS teamed up to optimize AI deployment at scale with new NVIDIA RTX PRO 4500 Blackwell GPUs in EC2 G7 instances, offering up to 4.6x AI inference performance.

Mercury Command is that for your finances.

Say what you need and the work gets done — payments, forecasting, categorization, invoices — across all of Mercury. You approve every action, and every answer is generated from your Mercury data with full account context and a traceable record.

TRY MERCURY COMMAND →

_*Mercury is a fintech company, not an FDIC-insured bank. Banking services provided through Choice Financial Group and Column N.A., Members FDIC._

SPACEX SIGNS COMPUTING POWER DEAL WITH OPEN-SOURCE AI STARTUP REFLECTION WORTH UP TO $6.3 BILLION

SpaceX signed a deal worth up to $6.3 billion with Reflection AI for access to its supercomputer Project Colossus. This agreement allows Reflection to use Nvidia GB300s for training open-source AI models, reflecting a shift toward increased demand for compute power in AI development. SpaceX aims to capitalize on this by offering its infrastructure to outside AI companies, bolstering its position in the AI and data center space.

OPENAI LAUNCHES NEW SECURITY TOOLS AND UPDATES GPT-5.5-CYBER

OpenAI recently launched an updated Codex Security plugin, GPT-5.5-Cyber (in limited release), a Daybreak Cyber Partner Program, and an open source security initiative called Patch the Planet. Daybreak is a defensive cyber stack for the AI era. OpenAI is promoting Daybreak through a partner model rather than direct broad model access. It aims to embed GPT-5.5 with Trusted Access for Cyber into existing security products and services while keeping access governed through partner systems.

ALIBABA'S AI VIDEO MODEL RISES TO NO. 2 IN GLOBAL RANKINGS, AS OPENAI'S SORA AND BYTEDANCE'S SEEDANCE FALL AWAY

Alibaba's HappyHorse 1.1 AI video generation model delivers production-ready video through an API built for integration into enterprise software stacks. It is now live on Alibaba Cloud Model Studio with a 40% sitewide launch discount for the first two weeks. HappyHorse supports text-to-video, image-to-video, and subject-to-video generation, as well as video editing. Its abilities cover the full spectrum of commercial video needs, from ideation through production to post-production.

GLM-5.2 RAISES THE BAR FOR OPEN MODELS

TheZvi examined GLM-5.2's capabilities and benchmark results, arguing it represents a significant improvement over previous open models. The analysis positioned it as one of the strongest openly available models while still trailing the leading frontier systems.

THE TEXT IN CLAUDE CODE'S “EXTENDED THINKING” OUTPUT IS NOT AUTHENTIC

Claude Code's 'Extended Thinking' reasoning is encrypted. Anthropic holds the key, and users' machines never receive it. The API returns a summary of the reasoning, and not the reasoning itself. Receiving the full thinking output requires an enterprise agreement.

MODEL SIZE SCALING IN 2023-2031

Targeting a particular speed of token generation puts a constraint on the total parameters of the model. If there isn't enough pretraining compute, models will remain smaller. This article looks at these considerations and estimates model sizes feasible for each year between 2023 and 2031. There are many assumptions that go into the estimates, which predict total parameters for models to reach 1.4 quadrillion in 2031.

KNOWLEDGE AGENTS: BEAT FRONTIER MODELS WITH BETTER STRUCTURE

Anthropic pulled its Mythos model, while the article's author developed smaller agentic models called "knowledge agents" to match larger frontier models. These agents enhance AI by injecting specific, relevant knowledge, performing well even with smaller models like Qwen 3.6 27B. The methodology involves embedding, structuring data, and multiple search passes, successfully augmenting LLMs for specialized queries and proprietary data.

GET STARTED WITH MODEL ROUTING →

MOEBIUS

Moebius is a highly efficient lightweight inpainting framework. The 0.22B model rivals and even surpasses the generation quality of the 11.9B industrial generalist FLUX.1-Fill-Dev. It does this with an under 15x acceleration in total inference time. Moebius liberates real-world image inpainting and AI object removal from parameter bloat.

USING CODEX FOR LONG-RUNNING PROJECTS

This guide outlines strategies for treating Codex as a persistent workspace that can maintain context across extended projects. It covers workflow management, task decomposition, and techniques for balancing autonomous execution with human oversight.

BONUS, FULLY REMOTE)

We're hiring a senior PMM to own product marketing at TLDR. You'll define our positioning, build out sales enablement, and lead every launch. Learn more [18].

ANTHROPIC SAYS CLAUDE MAY WANT TO SEE YOUR ID

Anthropic may require identity verification in certain circumstances starting on July 8. The company has not provided specific examples of when users will be asked to provide government-issued documents. An Anthropic spokesperson claims that the change will only apply to a small subset of users whose accounts are flagged but not outright banned. Anthropic uses Persona as its identity-checking provider.