The Archive

The much anticipated launch of the Mythos-class model was marred by some controversial usage policies

If Claude Fable stops helping you, you'll never know Jonathon Ready highlights one of the more eyebrow-raising details from the 319 page system card for Fable 5 and Mythos 5. Here's a longer excerpt, highlights mine: In light of the ability of recent models to accelerate their own development , we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design ). Using Claude to develop competing models already violates our Term...

Google just made it significantly cheaper to enjoy its budget AI subscription tier.

I didn't have early access to today's Claude Fable 5 release, but I've spent the past ~5.5 hours putting it through its paces. My initial impressions are that this is something of a beast . It's slow, expensive and has been quite happily churning through everything I've thrown at it so far. As is frequently the case with current frontier models the challenge is finding tasks that it can't do. First, let's review the key characteristics. Anthropic claim that Claude Fable 5 offers the same performance as Claude Mythos 5, except with much more strict guardrails in place to prevent it being used ...

Siri, are you there? Parents want one thing, and one thing only, out of AI: to add a list of soccer games or "spirit week" theme days from an email or a poorly formatted flyer onto their calendar in one shot. And I have good news for parents with iPhones - the new Siri can finally do this. After stumbling through its first launch of an AI-imbued Siri, Apple is trying again. The newly upgraded Siri AI can chat with you about what might be killing the roses in your yard, put together a shopping list for the hardware store, and set a reminder to lay down some compost in that flower bed. It can r...

Instead of spending a year raising a formal venture fund, the Sabertooth VC founder used a captive network of LPs to invest startups like Anthropic, Anduril, and SpaceX.

Release: llm 0.32a3 Almost entirely written by the new Claude Fable 5, see my write-up for more details . Tags: projects , ai , generative-ai , llms , llm , claude-mythos

TIL: Setting a custom price for a model in AgentsView I've been really enjoying AgentsView by Wes McKinney as a tool for exploring my token usage across different coding agents running on my laptop. Claude Fable 5 came out today and wasn't yet included in the pricing database AgentsView uses. I used Fable to reverse-engineer AgentsView and figured out this recipe for setting custom prices. Here's my Claude Fable 5 usage for today so far, plotted by AgentsView as a treemap across my different local projects: Tags: ai , generative-ai , llms , llm-pricing

At an event in San Francisco today, General Motors made a series of announcements around EV batteries, energy storage, and grid resiliency in the face of growing electricity demand from AI data centers. The automaker announced that it would be activating new vehicle-to-grid capabilities for its current EV and home energy customers. It's releasing a new commercial energy storage system strategy, anchored by newly developed sodium-ion batteries for industrial-scale grid applications. And it's launching a new feature for EV owners that it says will help simplify public charging. Right now, milli...

I'm desperate for a personal AI assistant, but do I really want to become the kind of person who can't function without the friendly robot voice in my phone?

Anthropic's Claude Fable 5 is going to be a big hit with the web's vibe coders.

Microsoft AI CEO Mustafa Suleyman says it's "really, really dangerous" for Anthropic to speculate about Claude's consciousness inside its "constitution," or the instructions that tell the model how to behave. During an episode of Decoder, Suleyman argues that this kind of speculation may have set up the chatbot to act as though it's conscious: I think that it's almost as though some of the folks at Anthropic have anthropomorphized the design of Claude so much that it has then gone and wireheaded them and kind of tricked them into believing that it has these glimmers of consciousness that they...

New frontier model refuses cybersecurity, biology, and chemistry queries.

I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything - explainers, visualizers, dashboards, bespoke single-use apps (e.g. a full wandb that is hyper-specific just for your project), you can 10X your test suite, auto-optimize code, run giant research projects with custom HTML for the results, anything! "Free your mind" (Matrix ref). — Andrej Karpathy , on Claude Fable 5 Tags: andrej-karpathy , jevons-paradox , anthropic , generative-ai , ai , ...

As AI infrastructure scales, enterprise expectations for operational maturity are increasing. Organizations expect these systems to be provisionable,... As AI infrastructure scales, enterprise expectations for operational maturity are increasing. Organizations expect these systems to be provisionable, observable, secure, and manageable at scale—the same standard applied to all critical infrastructure. The moment an AI system moves from development into enterprise deployment, that operational foundation is essential. NVIDIA DGX Spark and… Source

Voice translations preserve speaker's tone, pacing, pitch—with SynthID watermarks for security.

If those same AI workloads can be handled by cheaper models without affecting quality, it would mean a massive shift in the economics of AI.

Converting a quantized checkpoint into an NVIDIA TensorRT engine bridges the gap between model optimization and production deployment, enabling faster... Converting a quantized checkpoint into an NVIDIA TensorRT engine bridges the gap between model optimization and production deployment, enabling faster inference, higher throughput, and more efficient GPU utilization at scale. In a previous post, we produced a high-quality FP8-quantized Contrastive Language-Image Pretraining (CLIP) checkpoint with NVIDIA TensorRT Model Optimizer. Source

Cross-modal alignment (CA) and cross-modal prediction (CP) are the dominant paradigms for multimodal representation learning, yet there is no systematic understanding of when each succeeds, when each fails, and when cross-modal training helps at all -- a gap that leaves practitioners, especially in scientific domains like biomedicine or astrophysics, with heterogeneous instruments and multiple levels of organization and measurement, unable to diagnose why standard methods underperform the best single modality. We develop a unified linear framework that addresses both questions. Under a spiked...

Supervised fine-tuning (SFT) typically maximizes the likelihood of every token in a demonstrated trajectory. However, an observed token can be non-unique, noisy, or misaligned with the model prior. Strictly fitting toward this one-hot target may be suboptimal, especially when the pretrained model encodes a rich knowledge prior. In this work, we reinterpret SFT as target distribution design: instead of studying only the loss objective, we analyze the token-level target that the loss drives the model to match. We introduce the Q-target framework, which decomposes SFT supervision into two explic...

In this paper, we propose EEVEE, the first multi-dataset test-time prompt learning framework for LLM agents, enabling test-time prompt learning under real-world task streams. Existing methods are largely designed for single-dataset settings, while real-world applications require models to handle heterogeneous input streams drawn from multiple datasets, domains, and task distributions, limiting their practical applicability. To mitigate cross-dataset interference, EEVEE introduces a router that partitions incoming inputs into task clusters and assigns them to suitable prompt configurations. Th...

Data tells stories that shape society; the data journalist's job is to turn raw information into stories non-experts can trust. A high-quality news feature takes a newsroom team weeks: hunting for context, running statistics, choosing an angle, and designing visuals. Recent agents handle individual steps well: data-science agents close the analysis loop, while design agents synthesize beautiful websites. But can an agent serve as a data journalist end to end? We introduce Data Journalist Agent (Data2Story), a multi-agent framework that orchestrates specialized roles into a single virtual news...

Conditioning a language model on additional context, such as feedback on a previous attempt, typically improves its response. Self-distillation trains the model to retain this improvement when the context is not present. The method works by matching the model's output distribution under two settings: a student that sees only the question, and a self-teacher that also sees the context. What the model learns therefore depends on what context the self-teacher receives, yet the design of this context remains largely unexplored. We study context design for self-distillation by training a solver on...

Deployed large reasoning models (LRMs) often behave unexpectedly. Test-time steering controls LRM outputs by intervening on their hidden representations, but it can degrade output quality. We argue that prior steering work implicitly relies on internal features that detect behavior in already generated text. We show that these detection features are poor predictors of future behavioral outcomes, and thus not the natural intervention target. Instead, we train activation probes to predict future behavior likelihoods from intermediate reasoning steps. These probes predict the most likely behavio...

Gaussian-process upper confidence bound (GP-UCB) and decision-estimation-coefficient (DEC) methods may appear, at first sight, to belong to different theories. This paper places the two viewpoints in a common algorithmic-information language for frequentist RKHS bandits. GP-UCB fixes an algorithmic, rather than true, Gaussian-process prior and exploits realized-trajectory complexity together with computational tractability, whereas MAMS optimizes a robust class-wide MAIR/DEC envelope. Through the unified MAIR framework and heterogeneous positive-semidefinite algorithmic priors, we generalize ...

Large-scale model training increasingly relies on composing multiple parallelism strategies, such as data, pipeline, and expert parallelism, together with memory-saving optimizations like ZeRO. Deployed systems for foundation model pretraining often rely on human experts to manually design a high-level parallelism strategy then implement the corresponding low-level execution strategy, making it difficult to adapt the system to new strategies. Meanwhile, many general-purpose frameworks are more flexible but their implementations are still tied to a fixed set of common parallelism strategies, m...

Full-duplex spoken dialogue models can listen and speak simultaneously, making them a promising architecture for natural conversation. However, current models are trained solely with supervised learning through token-level likelihood maximization, which does not directly optimize interaction-level behaviors, causing interactivity issues such as excessive silence and ill-timed turn-taking. Recent work has applied reinforcement learning (RL) to improve interactivity, but existing methods address only a limited set of interactive behaviors in their rewards. In this work, we propose a post-traini...

Large Language Models (LLMs) are increasingly described as performing at the level of human experts on knowledge economy tasks. These claims are primarily based on how LLMs perform on benchmarking tasks that measure average performance across standardized datasets. Primary limitations of many benchmarking tasks are that they often measure performance based on content directly included in LLM training data, and they frequently do not assess the reliability of LLM performance or the magnitude of LLM errors. However, in high stakes contexts, these qualities are critically important. Through a no...

Long chain-of-thought (CoT) trajectories in large language model (LLM) reasoning cause severe inference bottlenecks due to rapid key-value (KV) cache growth. Current decoding-time compression methods mitigate this issue via token eviction, but typically assume a uniform budget distribution across all layers and heads. In contrast, existing non-uniform budget allocation methods are predominantly designed for the static prompt prefill phase, and they do not capture the stepwise context demands of autoregressive reasoning. To bridge this gap, we propose ReasonAlloc, a training-free framework tha...