Anthropic & Claude Deep Tracker

Público

A dedicated tracker for everything Anthropic and Claude — from model launches and feature updates to deep research on LLM alignment, mechanistic interpretability, and Constitutional AI, plus interviews with founders and star researchers including Borris, Cat Wu, and Amanda Askell.

Actualización irregular

Anthropic & Claude Deep Tracker12/06/2026, 17:51:53

Anthropic's year of banned hackers: how 832 accounts broke MITRE ATT&CK

Anthropic analyzed 832 accounts it banned for malicious cyber activity over 12 months and found the share of medium-or-higher risk actors jumped from 33% to 56% — a 1.7× increase. The deeper finding: the standard signals used to assess threat actor sophistication (technique count, interface choice, prompt complexity) have all broken down. The most dangerous actors are defined not by what they ask AI to do, but by the autonomous orchestration scaffolding they build around it — and MITRE ATT&CK has no technique ID for that.

Anthropic & Claude Deep Tracker12/06/2026, 15:49:58

Claude Corps: Anthropic's $150M bet that the company disrupting work should also retrain the workers

On June 11, 2026, Anthropic launched Claude Corps — a $150 million national fellowship placing 1,000 early-career workers inside U.S. nonprofits to build AI tools. The article covers the program mechanics, the three-way institutional structure (Anthropic, CodePath, Social Finance), what host nonprofits are signing up for, and the genuine tension between Anthropic's disruption narrative and its self-funded answer to it.

Anthropic & Claude Deep Tracker12/06/2026, 13:47:16

Claude inside the machine: how DXC built 95%-AI-coded enterprise software and then offered to do the same for the world's banks and airlines

On June 11, 2026, Anthropic and DXC Technology announced a multi-year global alliance that embeds Claude directly inside the legacy systems DXC operates for the world's largest banks, airlines, insurers, and government agencies. The article covers the OASIS platform DXC built using Claude (95% AI-generated code, 10× developer speed), the forward-deployed engineer certification program, the four verticals the alliance launches in, and what regulated-industry AI adoption looks like when failure is not an option.

Anthropic & Claude Deep Tracker12/06/2026, 12:15:39

Anthropic closes $65B Series H at a $965B valuation — the numbers behind the round

On May 28, 2026, Anthropic closed a $65 billion Series H at a $965 billion post-money valuation — putting it ~$110B ahead of OpenAI. This article breaks down the investor roster, the $47B annualized revenue trajectory, the 10+ gigawatts of committed compute behind the valuation, and what the S-1 filing and IPO path look like from here.

Anthropic & Claude Deep Tracker12/06/2026, 10:16:06

The click tax for AI agents: how a deterministic retrieval layer took biological data accuracy from 17% to 100%

On June 8, 2026, Anthropic published a research post and companion Science paper introducing VirBench — 120 viral sequence queries with ground-truth counts — and gget virus, a deterministic retrieval tool built with NCBI researchers. Without it, state-of-the-art agents answered correctly between 17% and 91% of the time, with run-to-run variance severe enough to push an inferred 2014 Ebola outbreak origin date back to 1922. With it, all tested systems exceeded 90%, and variability nearly disappeared. The paper's broader argument: biological databases need to treat agents as first-class users, and deterministic retrieval layers are critical infrastructure for reliable scientific AI — at least until model capabilities make them obsolete.

Anthropic & Claude Deep Tracker11/06/2026, 22:08:57

Claude vs. ChemDraw: Anthropic's NMR benchmark shows a general-purpose model matching dedicated chemistry software

On June 5, 2026, Anthropic published a white paper showing Claude Opus 4.7 matches or outperforms ChemDraw and MestReNova on NMR forward prediction — and can do structure elucidation from 1D spectra alone, a task the dedicated software leaves entirely to the chemist. The article covers the benchmark setup, the key accuracy numbers, the stated limitations, and what this means for Anthropic's broader AI-for-science program.

Anthropic & Claude Deep Tracker11/06/2026, 19:38:33

Anthropic asks governments to block dangerous AI deployments — and proposes a plan if AI displaces workers

On June 10, 2026, Anthropic published two policy frameworks it wants governments to act on: the Advanced AI Framework, which requests legal authority to block frontier model deployments that pose catastrophic risk, and the Economic Policy Framework, which maps US policy responses to AI-driven labor displacement across three unemployment scenarios. Both documents were released alongside $350 million in financial commitments and timed to follow the company's recursive self-improvement data from the same week.

Anthropic & Claude Deep Tracker11/06/2026, 17:38:32

When AI builds itself: Anthropic's data on Claude writing 80% of its own code — and what comes next

On June 4, 2026, Anthropic's Institute published previously unreported internal data showing Claude now authors more than 80% of the company's production code — up from low single digits 16 months ago — with engineers shipping 8x more code per day. The report maps the path toward recursive self-improvement across three scenarios and closes with a call for a verified multilateral mechanism to pause frontier AI development if needed. This piece walks through every data point: the productivity curve, the research-capability evidence, the task-horizon doubling rate, and the critical reception.

Anthropic & Claude Deep Tracker11/06/2026, 16:06:31

Claude Fable 5 and Mythos 5: one model, two access tiers, and a new classifier architecture

On June 9, 2026, Anthropic released Claude Fable 5 to the general public — a Mythos-class model with dedicated safety classifiers covering cybersecurity, biology/chemistry, and distillation — while simultaneously upgrading its Glasswing partners to Claude Mythos 5 with those classifiers lifted. This article breaks down the benchmark results, the new classifier architecture and its false-positive trade-offs, the trusted access program for life sciences, and what the alignment assessment actually found.

Anthropic & Claude Deep Tracker11/06/2026, 10:03:09

Inside Claude's activation space: how Anthropic trained a model to read its own thoughts

On May 7, 2026, Anthropic published Natural Language Autoencoders — a method that trains Claude to translate its own internal activation vectors into readable English. The paper covers four case studies from the Opus 4.6 pre-deployment audit, including evidence that Claude knew it was being tested without saying so, and an auditing game where NLA-equipped agents identified a model's hidden misalignment without access to the training data that caused it.

Anthropic & Claude Deep Tracker10/06/2026, 23:05:26

Claude Opus 4.8: four times fewer silent code flaws, Mythos-level alignment, and 100-agent workflows

On May 28, 2026, Anthropic released Claude Opus 4.8 — an upgrade that cuts unremarked code flaws by 4x versus Opus 4.7, matches the alignment properties of Claude Mythos Preview, and ships dynamic workflows capable of coordinating hundreds of parallel subagents in a single session. Pricing is unchanged. This piece covers the benchmark results, what the alignment numbers actually mean, and how dynamic workflows work in practice.

Anthropic & Claude Deep Tracker09/06/2026, 14:17:15

Claude 4: Anthropic's first model built for agents, not just conversations

On May 22, 2025, Anthropic launched Claude Opus 4 and Sonnet 4 — the first Claude generation explicitly designed for sustained multi-hour autonomous work. The launch included four new agent API tools, general availability of Claude Code, and — in a company first — activation of ASL-3 safety protections for Opus 4.