Augur Dispatch

Chain of evidence

Evidence for 2026-09-25

This frozen page shows Augur's claims and source links for one sent dispatch. Stored spot-checks appear only where the frozen edition supports them; absence is not presented as verification.

As of:

Bundle identity: evidence-bundle-v2-7ff20b780aee624951326c400388cfd34264fcc7651ed00a463a88b46dedbead

Format: evidence-bundle-v2 · 73 claims

Assertion 1

Black Forest Labs Latent Space Black Forest Labs Nate Jones So a small lab taking the top open slot from Nvidia looks like a shift in who leads.

Assertion status: No spot-check verdict is published for this assertion.

Black Forest Labs has made FLUX 3, a multimodal foundation model, available in early access.

Claim 35834 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 jointly learns from images, videos, and audio within a unified architecture.

Claim 35835 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 can generate video with audio up to 20 seconds in length in a single generation.

Claim 35836 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 was preferred over Luma Ray 3.2 in 93% of early comparative evaluations.

Claim 35839 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Mimic Robotics is a partner that has gained early access to FLUX 3 to develop FLUX-mimic.

Claim 35840 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Black Forest Labs plans to release open-weight access to the FLUX 3 Dev backbone model.

Claim 35842 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Black Forest Labs launched FLUX 3 Video, a unified multimodal model capable of text-to-video, image-to-video, video-to-video, and generative video-audio continuation, with native audio generation for all outputs.

Claim 36287 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Black Forest Labs announced FLUX 3-mimic, a Video-Action Model built on FLUX 3 for robot control and action prediction, currently being tested with Audi.

Claim 36288 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Black Forest Labs released FLUX 3 Action, an open-weight 7B world-action model derived from the FLUX 3 backbone, which achieves a state-of-the-art success rate on the RoboLab-120 leaderboard while using less than half the parameters of the previous best open model.

Claim 66821 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action achieves a 38.3% ± 0.38 success rate on the RoboLab benchmark, outperforming Cosmos 3 Nano, which achieved a 36.8% success rate.

Claim 66822 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action runs up to 3.95x faster than Cosmos 3 Nano in FP8 across consumer, workstation, and datacenter GPUs.

Claim 66823 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

A guidance-distilled version of FLUX 3 Action achieves a 42.2% ± 0.36 success rate on the RoboLab benchmark.

Claim 66824 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

The source asserts that GPT 6 Astra alone costs an average of $13.47 and takes 16 minutes per success, whereas FLUX 3 Action costs $0.09 and takes less than two minutes per success when running on an H200 GPU.

Claim 66826 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action utilizes a backbone less than half the size of Cosmos 3 Nano, enabled by multimodal Self-Flow pretraining and distillation that reduces sampling to a single step.

Claim 66828 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

NVIDIA builds tools for developers, including hardware, libraries, and models.

Claim 67080 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Nate Jones

No stored spot-check names this claim in this edition.

Cosmos is NVIDIA's world-model effort.

Claim 67081 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Nate Jones

No stored spot-check names this claim in this edition.

Ming-Yu Liu is the vice president of NVIDIA's Cosmos Lab.

Claim 67082 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Nate Jones

No stored spot-check names this claim in this edition.

NVIDIA wants to encourage the variety of applications built using world models.

Claim 67084 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Nate Jones

No stored spot-check names this claim in this edition.

Assertion 2

Black Forest Labs I do not know how that comparison was staged, and a general frontier model was never built for this job, so I read the ratio as a ceiling.

Assertion status: No spot-check verdict is published for this assertion.

Black Forest Labs released FLUX 3 Action, an open-weight 7B world-action model derived from the FLUX 3 backbone, which achieves a state-of-the-art success rate on the RoboLab-120 leaderboard while using less than half the parameters of the previous best open model.

Claim 66821 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action achieves a 38.3% ± 0.38 success rate on the RoboLab benchmark, outperforming Cosmos 3 Nano, which achieved a 36.8% success rate.

Claim 66822 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action runs up to 3.95x faster than Cosmos 3 Nano in FP8 across consumer, workstation, and datacenter GPUs.

Claim 66823 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

A guidance-distilled version of FLUX 3 Action achieves a 42.2% ± 0.36 success rate on the RoboLab benchmark.

Claim 66824 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

The source asserts that GPT 6 Astra alone costs an average of $13.47 and takes 16 minutes per success, whereas FLUX 3 Action costs $0.09 and takes less than two minutes per success when running on an H200 GPU.

Claim 66826 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action utilizes a backbone less than half the size of Cosmos 3 Nano, enabled by multimodal Self-Flow pretraining and distillation that reduces sampling to a single step.

Claim 66828 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Assertion 3

Sakana AI For an engineering lead running a manipulation pilot, the budget question becomes which model absorbs thousands of retries, and my answer is the cheap open one, with the frontier model kept as a judge or a generator of training data.

Assertion status: No spot-check verdict is published for this assertion.

Jürgen Schmidhuber has officially joined Sakana AI as Chief Scientific Advisor, a role he holds alongside his current positions.

Claim 66952 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Sakana AI describes Jürgen Schmidhuber as universally recognized as the father of modern AI.

Claim 66953 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Jürgen Schmidhuber will be involved in Sakana AI's newly formed Recursive Self-Improvement (RSI) Lab.

Claim 66954 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Sakana AI is developing Agent-Native World Models that can simulate the physical consequences of actions before they occur.

Claim 66955 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Jürgen Schmidhuber’s foundational vision from his 1987 thesis on recursive self-improvement and meta-learning has directly shaped Sakana AI's recent breakthroughs, including Darwin Gödel Machine and The AI Scientist.

Claim 66956 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Sakana AI aims to use World Models to optimize supply chains and robotic deployments for Japanese industry without costly real-world trial and error.

Claim 66957 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Jürgen Schmidhuber traveled to Tokyo regularly to engage with Sakana AI's team and support the broader Japanese AI ecosystem.

Claim 66958 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

The ultimate objective of Sakana AI's RSI Lab is to trigger a compounding cycle of scientific discovery aimed at improving machine intelligence.

Claim 66959 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Assertion 4

Hugging Face Blog Apple Machine Learning Research Ollama Releases These are three pieces of one shift: the same laptop gets faster without new hardware.

Assertion status: No spot-check verdict is published for this assertion.

Apple's system-wide dictation on-device speech transcription uses a tokenizer encoder that maps waveform windows to representations for a sparsely activated foundation model using Instruction-Following Pruning.

Claim 67000 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Apple Machine Learning Research

No stored spot-check names this claim in this edition.

Because the foundation model is sparsely activated, only a small subset of its experts occupies DRAM at any time, causing the always-on tokenizer to compete for memory.

Claim 67001 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Apple Machine Learning Research

No stored spot-check names this claim in this edition.

The authors compressed a streaming neural audio tokenizer by distilling it to regress the teacher model's per-frame pre-quantizer latent representation using a squared-error objective.

Claim 67002 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Apple Machine Learning Research

No stored spot-check names this claim in this edition.

The LiquidAI LFM2.5-VL-DSpark drafter adds 280M parameters, representing an 8.9% increase over the 3B target model.

Claim 67092 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

LiquidAI reports that the LFM2.5-VL-DSpark drafter achieves up to 3.13x faster decoding on an M5 Max device.

Claim 67093 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

LiquidAI reports up to 2.66x faster decoding for the LFM2.5-VL-DSpark drafter on an H100 GPU.

Claim 67094 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

LiquidAI states that the LFM2.5-VL-DSpark drafter provides day-one integration support for llama.cpp, MLX-VLM, and SGLang.

Claim 67095 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

In the Ollama v0.40.0-rc0 release, model architectures supported by the MLX runner run by default on Apple Silicon devices.

Claim 67381 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

During the release candidate period for Ollama v0.40.0, additional models will be tested and enabled.

Claim 67382 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

Assertion 5

Hugging Face Blog

Assertion status: No spot-check verdict is published for this assertion.

Liquid AI claims that its LFM2.5-DSpark technology provides up to 3.2x faster inference speeds.

Claim 49687 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Liquid AI reports that the LFM2.5-DSpark draft models achieve a 3.18x throughput improvement on an NVIDIA H100 GPU.

Claim 49688 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Liquid AI reports that the LFM2.5-DSpark draft models achieve a 2.87x throughput improvement on an M4 Max MacBook Pro.

Claim 49689 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Liquid AI states that the LFM2.5-DSpark draft models cut function-calling latency by an average of 57% for the LFM2.5-2.6B model.

Claim 49690 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Liquid AI reports that for the LFM2.5-1.2B-Instruct model, speedup varies by as much as 52% depending on the underlying text distribution.

Claim 49696 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Liquid AI states that the DSpark draft models are approximately 300M parameters in size.

Claim 49698 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Assertion 6

Google DeepMind Blog Google DeepMind Blog The same day Google Research published a multi-agent director for long-form video built on Gemini and Veo, which inherits SynthID, Google's invisible watermark, from those models.

Assertion status: No spot-check verdict is published for this assertion.

Google DeepMind introduced two new AI models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, on September 15, 2026.

Claim 61822 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live processes visual inputs in near real-time to enrich conversational context.

Claim 61829 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

All audio generated by Google DeepMind's AI products is watermarked with SynthID to ensure detectability.

Claim 61831 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live became available to developers via the Gemini API and Google AI Studio on September 15, 2026.

Claim 61832 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live became available to enterprises in private preview via Gemini Enterprise on September 15, 2026.

Claim 61833 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live became available to the general public in Search Live on September 15, 2026.

Claim 61834 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live Extended Thinking became available to developers via the Gemini API and Google AI Studio on September 15, 2026.

Claim 61835 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live Extended Thinking became available to enterprises in private preview via Gemini Enterprise on September 15, 2026.

Claim 61836 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live Extended Thinking became available to Google AI Pro and Ultra subscribers in Workspace on September 15, 2026.

Claim 61837 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Google DeepMind launched Gemini 3.8 Live with Live Avatar, a feature that pairs near real-time video generation with speech to create a dynamic visual persona for native live dialogue models.

Claim 67235 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

The Live Avatar feature enables enterprises to expand virtual offerings by listening, seeing, and speaking with a visual persona that supports precise lip-syncing, natural expressions, and fluid turn-taking.

Claim 67236 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live with Live Avatar became available in Gemini Enterprise starting on September 24, 2026.

Claim 67237 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

The Live Avatar feature processes visual and audio inputs simultaneously to generate enriching conversations for enterprise agents.

Claim 67238 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Live Avatar uses asynchronous tool calling to trigger tool calls and fetch data in the background while continuing active dialogue, ensuring uninterrupted conversational flow.

Claim 67239 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Live Avatar features native multilingual speech-to-speech synchronization that dynamically adapts lip-sync and expressions across 97 languages without degrading video fidelity or introducing visual drift.

Claim 67240 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

All output generated by Gemini's AI products is watermarked with SynthID to help ensure AI-generated content remains detectable and minimize misinformation.

Claim 67242 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Assertion 7

Google Research Blog That paper shows Google building the video plumbing underneath features like the avatar; it is not evidence that any buyer wants a face on a support agent.

Assertion status: No spot-check verdict is published for this assertion.

Google researchers introduced a unified multi-agent framework that autonomously generates temporally consistent, long-form video narratives.

Claim 67314 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google Research Blog

No stored spot-check names this claim in this edition.

The AI video co-director framework is built as an orchestration layer on top of Google's Gemini and Veo models.

Claim 67315 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google Research Blog

No stored spot-check names this claim in this edition.

The AI video co-director framework natively inherits safety mechanisms such as SynthID watermarking from its underlying models.

Claim 67316 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google Research Blog

No stored spot-check names this claim in this edition.

Assertion 8

Princeton CITP Put together, they say the return rests less on construction cost than on how long the expensive part stays useful.

Assertion status: No spot-check verdict is published for this assertion.

Goldman Sachs estimates that compute equipment inside AI data centers costs 3-4 times more than the physical data center structure itself.

Claim 67329 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Princeton CITP - Freedom to Tinker

No stored spot-check names this claim in this edition.

Hyperscalers peg the useful life of their servers at 5 to 6 years.

Claim 67330 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Princeton CITP - Freedom to Tinker

No stored spot-check names this claim in this edition.

The Anthropic Institute's analysis posits that AI's net contribution to U.S. GDP in 2030 could be 1.6% under modest assumptions, 8.3% under substantial assumptions, and 32.4% under extreme assumptions.

Claim 67331 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Princeton CITP - Freedom to Tinker

No stored spot-check names this claim in this edition.

Assertion 9

Nate Jones The two figures measure different things, a chip and the box it sits in, so I read this as a stretched assumption rather than a proven one.

Assertion status: No spot-check verdict is published for this assertion.

Conventional assumptions for GPU life are 3 to 5 years. (Fact, 2026-08-16)

Claim 47783 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Nate Jones

No stored spot-check names this claim in this edition.

Assertion 10

Princeton CITP If you are the executive signing a multi-year AI commitment, you are implicitly betting on the vendor's depreciation schedule, and I would ask for that assumption in writing before discussing price.

Assertion status: No spot-check verdict is published for this assertion.

Goldman Sachs estimates that compute equipment inside AI data centers costs 3-4 times more than the physical data center structure itself.

Claim 67329 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Princeton CITP - Freedom to Tinker

No stored spot-check names this claim in this edition.

Hyperscalers peg the useful life of their servers at 5 to 6 years.

Claim 67330 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Princeton CITP - Freedom to Tinker

No stored spot-check names this claim in this edition.

The Anthropic Institute's analysis posits that AI's net contribution to U.S. GDP in 2030 could be 1.6% under modest assumptions, 8.3% under substantial assumptions, and 32.4% under extreme assumptions.

Claim 67331 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Princeton CITP - Freedom to Tinker

No stored spot-check names this claim in this edition.

Assertion 11

- Black Forest Labs has promised FLUX 3 Dev as an open-weight variant since July; a weights drop would put the base model in outside teams' hands, though FLUX 3 Action is a separate model and a Dev release alone would not let anyone rerun its RoboLab numbers. Black Forest Labs

Assertion status: No spot-check verdict is published for this assertion.

Black Forest Labs has made FLUX 3, a multimodal foundation model, available in early access.

Claim 35834 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 jointly learns from images, videos, and audio within a unified architecture.

Claim 35835 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 can generate video with audio up to 20 seconds in length in a single generation.

Claim 35836 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 was preferred over Luma Ray 3.2 in 93% of early comparative evaluations.

Claim 35839 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Mimic Robotics is a partner that has gained early access to FLUX 3 to develop FLUX-mimic.

Claim 35840 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Black Forest Labs plans to release open-weight access to the FLUX 3 Dev backbone model.

Claim 35842 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Assertion 12

- Ollama says more models will be tested and enabled during the v0.40.0 release candidate period, so the stable release will show how much of the local catalog runs on MLX by default. Ollama Releases

Assertion status: No spot-check verdict is published for this assertion.

In the Ollama v0.40.0-rc0 release, model architectures supported by the MLX runner run by default on Apple Silicon devices.

Claim 67381 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

During the release candidate period for Ollama v0.40.0, additional models will be tested and enabled.

Claim 67382 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

Assertion 13

- Google's Live Avatar is now in Gemini Enterprise; named customer deployments, rather than launch demos, would show whether buyers want a face on their agents. Google DeepMind Blog

Assertion status: No spot-check verdict is published for this assertion.

Google DeepMind launched Gemini 3.8 Live with Live Avatar, a feature that pairs near real-time video generation with speech to create a dynamic visual persona for native live dialogue models.

Claim 67235 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

The Live Avatar feature enables enterprises to expand virtual offerings by listening, seeing, and speaking with a visual persona that supports precise lip-syncing, natural expressions, and fluid turn-taking.

Claim 67236 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Gemini 3.8 Live with Live Avatar became available in Gemini Enterprise starting on September 24, 2026.

Claim 67237 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

The Live Avatar feature processes visual and audio inputs simultaneously to generate enriching conversations for enterprise agents.

Claim 67238 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Live Avatar uses asynchronous tool calling to trigger tool calls and fetch data in the background while continuing active dialogue, ensuring uninterrupted conversational flow.

Claim 67239 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Live Avatar features native multilingual speech-to-speech synchronization that dynamically adapts lip-sync and expressions across 97 languages without degrading video fidelity or introducing visual drift.

Claim 67240 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

All output generated by Gemini's AI products is watermarked with SynthID to help ensure AI-generated content remains detectable and minimize misinformation.

Claim 67242 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Google DeepMind Blog

No stored spot-check names this claim in this edition.

Assertion 14

- Sakana's Recursive Self-Improvement Lab, its effort to have AI systems improve AI research, now has Schmidhuber attached, and a paper or demo of its Agent-Native World Models would turn a stated plan into something testable. Sakana AI

Assertion status: No spot-check verdict is published for this assertion.

Jürgen Schmidhuber has officially joined Sakana AI as Chief Scientific Advisor, a role he holds alongside his current positions.

Claim 66952 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Sakana AI describes Jürgen Schmidhuber as universally recognized as the father of modern AI.

Claim 66953 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Jürgen Schmidhuber will be involved in Sakana AI's newly formed Recursive Self-Improvement (RSI) Lab.

Claim 66954 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Sakana AI is developing Agent-Native World Models that can simulate the physical consequences of actions before they occur.

Claim 66955 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Jürgen Schmidhuber’s foundational vision from his 1987 thesis on recursive self-improvement and meta-learning has directly shaped Sakana AI's recent breakthroughs, including Darwin Gödel Machine and The AI Scientist.

Claim 66956 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Sakana AI aims to use World Models to optimize supply chains and robotic deployments for Japanese industry without costly real-world trial and error.

Claim 66957 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Jürgen Schmidhuber traveled to Tokyo regularly to engage with Sakana AI's team and support the broader Japanese AI ecosystem.

Claim 66958 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

The ultimate objective of Sakana AI's RSI Lab is to trigger a compounding cycle of scientific discovery aimed at improving machine intelligence.

Claim 66959 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Sakana AI

No stored spot-check names this claim in this edition.

Assertion 15

- Fireworks ships ARCv3, a tool that shrinks the model updates sent between datacenters during training, as a package for teams running their own trainers, and adoption outside its own platform would test the cross-region training claim. Fireworks AI

Assertion status: No spot-check verdict is published for this assertion.

Fireworks' ARCv3 compressor reduces the size of BF16 weight update payloads sent from trainers to reinforcement learning rollout machines by nearly 50% compared to the previous ARCv2 version.

Claim 66809 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

Across a benchmark of 1,000 production RL weight-update deltas, ARCv3 achieved an average compressed payload size of approximately 0.19% of the original BF16 weight size.

Claim 66810 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

ARCv3 is available in the Fireworks Training API via the fireworks-delta-compression package for teams using their own trainers with Fireworks rollouts.

Claim 66811 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

ARCv3 is the default compressor for rollout fleets when users employ the Fireworks Trainer SDK.

Claim 66812 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

ARCv3 produces smaller delta payloads than PULSESync (0.35%), ARCv2 (0.36%), and Hugging Face TRL's delta-compression implementation (0.67%) in benchmarks of BF16 weight tensors.

Claim 66813 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

ARCv3 encodes weight update asymmetry by omitting unchanged values, contributing only mantissa bits for common mantissa-only changes, and packing rarer exponent or sign flips separately.

Claim 66815 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

For each tensor, ARCv3 runs several general-purpose compression algorithms in parallel and retains the one producing the smallest output.

Claim 66816 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

The trainer uploads compressed weight delta slices to shared object storage in parallel, after which the Fireworks API signals regions to pull pieces and reconstruct the model locally.

Claim 66818 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

ARCv3 helps keep a globally distributed rollout fleet closer to the current policy, which enables cross-region reinforcement learning to work at scale.

Claim 66820 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Fireworks AI

No stored spot-check names this claim in this edition.

Assertion 16

Black Forest Labs The FLUX 3 Dev open-weight backbone remains a promise dating to July rather than a shipped release.

Assertion status: No spot-check verdict is published for this assertion.

Black Forest Labs released FLUX 3 Action, an open-weight 7B world-action model derived from the FLUX 3 backbone, which achieves a state-of-the-art success rate on the RoboLab-120 leaderboard while using less than half the parameters of the previous best open model.

Claim 66821 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action achieves a 38.3% ± 0.38 success rate on the RoboLab benchmark, outperforming Cosmos 3 Nano, which achieved a 36.8% success rate.

Claim 66822 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action runs up to 3.95x faster than Cosmos 3 Nano in FP8 across consumer, workstation, and datacenter GPUs.

Claim 66823 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

A guidance-distilled version of FLUX 3 Action achieves a 42.2% ± 0.36 success rate on the RoboLab benchmark.

Claim 66824 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

The source asserts that GPT 6 Astra alone costs an average of $13.47 and takes 16 minutes per success, whereas FLUX 3 Action costs $0.09 and takes less than two minutes per success when running on an H200 GPU.

Claim 66826 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 Action utilizes a backbone less than half the size of Cosmos 3 Nano, enabled by multimodal Self-Flow pretraining and distillation that reduces sampling to a single step.

Claim 66828 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Assertion 17

Black Forest Labs

Assertion status: No spot-check verdict is published for this assertion.

Black Forest Labs has made FLUX 3, a multimodal foundation model, available in early access.

Claim 35834 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 jointly learns from images, videos, and audio within a unified architecture.

Claim 35835 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 can generate video with audio up to 20 seconds in length in a single generation.

Claim 35836 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

FLUX 3 was preferred over Luma Ray 3.2 in 93% of early comparative evaluations.

Claim 35839 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Mimic Robotics is a partner that has gained early access to FLUX 3 to develop FLUX-mimic.

Claim 35840 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Black Forest Labs plans to release open-weight access to the FLUX 3 Dev backbone model.

Claim 35842 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Black Forest Labs

No stored spot-check names this claim in this edition.

Keep exploring

Move on to Augur's other sections.