Claude Science Completes Two Years' Work in a Few Weeks, Is 10x Research Acceleration Really Here?

marsbitPublicado em 2026-07-01Última atualização em 2026-07-01

Resumo

Claude Science, a new AI workbench from Anthropic, is being tested by scientists, reportedly accelerating specific research workflows by up to 10x. A neuro-scientist at the Allen Institute completed a lengthy literature review in weeks instead of nearly two years using the tool, which automates tasks like citation verification. The platform is an integrated environment for macOS and Linux, connecting to local or remote computing resources. It streamlines the fragmented research process—literature analysis, computation, visualization, and drafting—into a single, auditable workflow. A key feature is its emphasis on reproducibility: every chart generated includes the exact code, environment, and history used to create it. Claude Science uses a multi-agent system. A coordinator manages over 60 pre-configured skills for life sciences (genomics, proteomics, etc.) and can spawn specialized agents. A dedicated reviewer agent checks citations and calculations for accuracy, creating a form of internal AI peer review. The system operates with a human-in-the-loop, requiring user approval for major steps. Initial applications are in life sciences. Examples include target identification for biotech company Manifold Bio and germline variant analysis for glioma research at UCSF, completing analyses in roughly one-tenth the previous time. The approach contrasts with competitors: Google focuses on proprietary models like AlphaFold, while OpenAI is advancing models' scientific reasoning wit...

Two years' work, now completed in a few weeks.

Recently, neuroscientist Jérôme Lecoq and his team at the Allen Institute compressed the writing time of a long-form review article from nearly two years to just a few weeks.

Jérôme Lecoq had a backlog of about 10 reviews, many exceeding 100 pages, with every single citation checked sentence by sentence by an AI agent.

The tool helping him was Claude Science, the new application just launched by Anthropic.

On June 30, 2026, Anthropic released Claude Science, positioned as an AI workbench for scientists. (Image source: Anthropic official blog)

According to Anthropic, this task would have taken the scientist and his team two years in the past.

Anthropic's positioning for Claude Science is not just a smarter research model, but an AI workbench tailored for scientists.

Its true breakthrough lies in: for the first time, breaking down scientific research into an auditable, step-by-step pipeline.

Currently, Claude Science is in beta on macOS and Linux, available to Pro, Max, Team, and Enterprise users.

What's Really Changing is the Entire Research Toolchain

Anyone who has done research understands the drudgery:

A project requires jumping between dozens of databases, each with its own schema and query language;

File formats are all over the place, requiring custom pipelines and viewers for each;

You have a row of tools on hand: PubMed for literature, Jupyter for code, R for statistics, cluster terminals for submitting jobs...

Constant context switching, leaving little time for actual scientific thinking amidst the work of moving, splicing, and debugging.

What Claude Science does is bundle these fragmented scenarios into a single execution environment:

Literature analysis, multi-step computation, chart polishing, manuscript drafting—all stages are completed in the same environment, so you don't interrupt your train of thought switching tools.

It can run on your local macOS or Linux machine, connect via SSH to remote machines, or attach to a high-performance computing (HPC) login node.

Just like using Jupyter normally, it goes where the data is.

It even handles computational resource scheduling.

Large tasks like protein folding or running a genomic pipeline on massive data used to require researchers to babysit: setting up jobs, queuing for cluster time, monitoring success or failure, and pulling results back—half a day gone in a flash.

Claude Science takes over this flow: drafting a plan, asking for permission before touching new resources, letting you review or revert tasks before they are written and submitted, scaling analysis from 1 GPU to hundreds.

Claude Science dispatches an 8-group scVI hyperparameter scan to a lab's A100 cluster. The Notebook on the right shares the same live kernel with the agent, with variables and state synchronized in real-time. (Image source: Anthropic official blog)

More importantly, sensitive data doesn't leave the original system; only the context truly needed for each step is sent to Claude.

Every Chart Comes with Traceable Code

Science inherently deals with visuals: protein 3D structures, genome browser tracks, chemical formulas—these are essentially diagrams.

Building on this, Claude Science, while generating charts and drafts, also outputs the code that created them and can render them natively.

The key lies in reproducibility.

Whenever Claude Science generates a chart, it "pins" the exact code that created it, along with the runtime environment, plain-language descriptions, and the full conversation history, right onto the chart.

On the left, a cell chart across 138 species; on the right, the exact code that generated it is displayed side-by-side. Annotating with a sentence can have the agent modify the chart. Every result is reproducible and traceable to its code. (Image source: Anthropic official blog)

Months often pass from a paper's submission to publication; months later, when a reviewer asks you to rerun a specific chart, you can easily reproduce the entire chain of inputs, process, and results on the spot.

Want to edit the chart? Just speak—"remove gridlines," "switch the y-axis to log," the agent directly edits the code it wrote.

You can also fork the session at any point to try two different approaches simultaneously, without disrupting the original thread.

In short, research is integrated for the first time into an auditable workflow, with code, environment, and history all placed within a closed loop.

One Agent Writes, Another Specializes in Finding Errors

Behind Claude Science, it's not a single agent working alone.

You're facing a coordinating agent that manages over 60 pre-configured skills and connectors for genomics, single-cell analysis, proteomics, structural biology, and cheminformatics.

When work piles up, it can spawn additional agents for division of labor and can call upon expert agents you've created yourself.

The most clever part is the reviewer agent.

It specializes in checking citations and calculations, hunting down wrong citations, unsourced numbers, charts that don't match the code, marking them, and fixing them itself.

In the Allen Institute case, the team used an actor-critic pairing: one agent responsible for writing, another specialized in evaluating its accuracy and the veracity of citations.

This structure already hints at a prototype for "AI internal peer review."

But one boundary must be clear: it's human-in-the-loop throughout.

Before utilizing new resources, it seeks authorization; every decision you can review and revoke. It automates the process, not the scientific discovery itself.

It also integrates with NVIDIA's BioNeMo Agent Toolkit, connecting natively to life science models like Evo 2, Boltz-2, and OpenFold3.

Models, data, and pipelines your lab trusts can be saved as reusable skills and integrated, automatically inherited by future sessions.

Claude Science's First Stop is Life Sciences

Claude Science's initial focus is on life sciences.

Genomics, single-cell, proteomics, structural biology, cheminformatics—ready to use out of the box.

It can read literature, query 60+ scientific databases like UniProt, PDB, Ensembl, ClinVar, ChEMBL, and GEO—you no longer need to learn how to use each disparate database one by one.

Claude Science comes with pre-configured environments for genomics, single-cell, proteomics, and cheminformatics, backed by 60+ scientific databases. (Image source: Anthropic official blog)

Manifold Bio works on tissue-targeting drugs.

They use Claude Science to nominate targets for the latest experiments, evaluating surface expression, transport, and safety for each tissue and target, then ranking candidates according to standards learned from their own proprietary data.

Manifold says ordinary programming assistants can't do this; Claude Science can complete it end-to-end, getting the right data, making the right judgments, and carrying context from past projects.

There are even more hardcore examples.

An epidemiology associate professor at the UCSF Brain Tumor Center uses it for molecular epidemiology studies of glioma, analyzing how thousands of small-effect germline variants combine to shape individual susceptibility.

According to Anthropic, this germline analysis was completed by Claude Science in about 1/10th the time, and his team independently verified the results, confirming both speed and reliability.

However, these 10x acceleration scenarios are currently limited to review writing, genomic analysis, and specific pipeline automation, and do not equate to "10x acceleration of research overall."

Simultaneously, the threshold for research credibility is being redefined.

In the past, the reliability of a study depended on peer review and whether others could reproduce it.

Reproducibility has long been a major pain point in science—code gets lost, environments change, months later even the authors themselves can't reproduce that original chart.

Claude Science ensures every chart has traceable code, every result is linked to its environment and history. It might be the first to truly cross the reproducibility hurdle.

Three Players on the Same Track

In the bio-research track, the three giants are all competing, each with a different approach.

Google bets on proprietary models, OpenAI bets on the model's scientific IQ, while Anthropic bets on workflow.

Google holds exclusive models like AlphaFold and AlphaGenome that others don't have, going directly to market.

OpenAI takes a different path.

In April this year, it launched GPT-Rosalind, a cutting-edge model specifically built for biological reasoning and drug discovery.

Now it's going further, training the model's "scientific judgment."

It just launched GeneBench-Pro, specifically testing if models can make judgments like a computational biologist: 129 questions, spanning genomics, population genetics, all the way to clinical diagnosis, specifically testing the feel for "does the data support this question" and "which step should be redone."

The strongest model, GPT-5.6 Sol, scored 28.7%, 31.5% with Pro mode; GPT-5 from a few generations ago was less than 5%.

OpenAI itself says that at this rate, the benchmark could be topped by the end of the year.

But even the strongest models only solve less than a third. The unsolvable portion is precisely where human scientists remain.

The AI shortcomings exposed by GeneBench-Pro are also obvious:

Models can start but can't close the final loop, like whether to exclude a batch of anomalous data, how to change course when a hypothesis is overturned—these judgments still require the scientist's own decision.

Claude Science doesn't bypass this either; proposals are submitted for human review, every decision left for human revocation. It automates the process, judgment is not handed to the model, humans remain in the loop.

For scientists like Lecoq, whether a review is reproducible, whether it still holds up months later, matters more than an extra few percentage points on a leaderboard.

Claude Science is betting precisely on making AI research truly land in the daily routine of the lab.

References:

https://www.anthropic.com/news/claude-science-ai-workbench

https://openai.com/index/introducing-genebench-pro/

This article is from the WeChat public account "New Zhiyuan," author: ASI Apocalypse

Perguntas relacionadas

QWhat specific scientific task was dramatically accelerated by Claude Science according to the article?

AThe article describes how Claude Science enabled neuroscientist Jérôme Lecoq and his team to complete the writing of a long-form scientific review in a few weeks, a task that would have taken nearly two years previously.

QWhat is the core difference between Claude Science and other AI models for science like Google's or OpenAI's, as stated in the article?

AThe article states that while Google bets on proprietary models (e.g., AlphaFold) and OpenAI focuses on a model's research intelligence (e.g., GPT-Rosalind), Anthropic's Claude Science bets on workflow. It is positioned as an AI workbench that integrates and automates the entire scientific toolchain into an auditable pipeline, not just a smarter model.

QHow does Claude Science address the critical challenge of reproducibility in scientific research?

AClaude Science addresses reproducibility by packaging every generated figure with the exact code that produced it, the runtime environment, a natural language description, and the full conversation history. This creates a traceable, closed-loop workflow where results can be easily recreated and audited later.

QDescribe the role of the 'reviewer agent' within the Claude Science system.

AAccording to the article, the 'reviewer agent' is a specialized component that checks citations and computations. It identifies incorrect citations, numbers without sources, and figures that don't match the code, then flags and corrects them. This creates a system resembling 'internal AI peer review'.

QWhat limitation of current AI models in scientific research was highlighted by OpenAI's GeneBench-Pro benchmark, as mentioned in the article?

AOpenAI's GeneBench-Pro benchmark highlighted that while models can start analyses, they struggle with the final, critical judgment calls. This includes decisions like whether to exclude a batch of outlier data or how to change the research approach when a hypothesis is disproven. These core scientific judgments still require a human scientist's decision.

Leituras Relacionadas

Goldman Sachs In-Depth Report: Who Will Be the Long-Term Winners in China's AI Large Model Industry?

Goldman Sachs Report: China's AI Models at an Inflection Point China's open-source/open-weight large language models (LLMs) have reached performance parity with top global proprietary models, according to a Goldman Sachs report. This is driven by architectural innovations and higher parameter efficiency, allowing Chinese models to achieve comparable capabilities at 2%-10% the parameter size and significantly lower cost. The market is evolving into a two-tiered structure: a high-end segment (e.g., GLM5.2, Qwen3.7 Max) with premium pricing and a low-end, price-sensitive segment for global SMEs and individual users. Key points: * **Cost & Performance:** Innovations like Mixture of Experts (MoE) enable high performance with smaller models. Projects like Meituan's LongCat 2.0, trained on domestic hardware, highlight progress in tech self-sufficiency. * **Open-Source Strategy:** Most Chinese players use open-source/open-weight models for flexibility and ecosystem growth. However, Goldman notes this may underreport actual deployment and revenue. A shift toward "open-weight + community license" models with revenue sharing (e.g., MiniMax) could improve monetization. * **Market Shift & Global Expansion:** Enterprise AI adoption is shifting from "token maximization" to "ROI-first." International expansion, especially in non-US markets, is a major growth driver. Chinese models are increasingly available on global platforms like AWS Bedrock and Microsoft Copilot. * **Competitive Landscape:** Using a framework based on pricing power, cost advantage, and financial strength, Goldman identifies **Zhipu AI and DeepSeek** as the strongest in foundational text models, and **ByteDance** as the leader in multimodal/video generation. The report maintains Buy ratings on MiniMax and Kuaishou. * **Market Growth:** China's AI model API and subscription revenue is projected to grow from an estimated ¥35 billion in 2026 to ¥879 billion by 2030.

marsbitHá 10m

Goldman Sachs In-Depth Report: Who Will Be the Long-Term Winners in China's AI Large Model Industry?

marsbitHá 10m

Goldman Sachs Deep Dive Report: Who Will Become the Long-Term Winners in China's AI Large Model Industry?

Goldman Sachs Report: Who Will Be the Long-Term Winners in China's AI Large Model Industry? China's AI large model sector is at a historic inflection point. Goldman Sachs argues that the intelligence of Chinese open-source/open-weight models is approaching top global proprietary models. Rapid adoption by domestic enterprises and global SMEs is creating a data flywheel effect that will further drive model iteration. The evolution is summarized as moving from "DeepSeek's cost-efficiency moment last year to GLM's model-intelligence moment this year." Chinese models achieve near-state-of-the-art performance at significantly lower cost, primarily due to architectural innovations like Mixture of Experts (MoE) and higher parameter efficiency. Models like DeepSeek V4 Pro (1.6T params), GLM5.2 (0.7T), and MiniMax M3 (0.4T) are much smaller than global leaders. Recent advancements in coding capability are attributed to better data curation and RLHF. Landmarks like Meituan's LongCat 2.0, trained fully on domestic AI chips, demonstrate progress in hardware stack independence. The market is forming a "two-tiered structure." The high-end tier (e.g., GLM5.2, Alibaba's Qwen3.7 Max) prices around $1 per million tokens, about 10-25% of US top models, with estimated inference gross margins of 10-20%. The low-end tier (priced as low as $0.06-$0.2 per million tokens) targets price-sensitive global SMEs and individuals. MiniMax derives 60-70% of revenue overseas. Goldman forecasts China's AI model API/subscription revenue to grow from an estimated RMB 35bn in 2026 to RMB 879bn by 2030. Most Chinese players adopt open-source/open-weight strategies for deployment flexibility and community feedback, though this limits monetization as deployments on third-party platforms (e.g., Alibaba Cloud) may not generate direct revenue. A shift towards "open-weight + community license" models with revenue-sharing agreements (like MiniMax's approach) could improve unit economics. International expansion, particularly in non-US markets, is the key growth driver. The global enterprise AI paradigm is shifting from "token maximization" to "ROI prioritization." Chinese models are already hosted on major global platforms like AWS Bedrock and are under consideration for integration into Microsoft Copilot. Using a competitive framework based on pricing power, cost advantage, and financial strength, Goldman identifies the strongest players: In foundational text models, Zhipu AI (initiated coverage) and DeepSeek lead. In multimodal/video generation, ByteDance's Seed is the frontrunner, with Kuaishou's Kling and MiniMax's Hailuo also well-positioned. Goldman maintains a Buy rating on MiniMax, citing its attractive valuation.

链捕手Há 14m

Goldman Sachs Deep Dive Report: Who Will Become the Long-Term Winners in China's AI Large Model Industry?

链捕手Há 14m

Circle secures final OCC approval for national trust bank to strengthen USDC infrastructure

Circle has received final approval from the U.S. Office of the Comptroller of the Currency (OCC) to establish Circle National Trust, a federally regulated national trust bank. This marks a key regulatory milestone, bringing core parts of the company's USDC stablecoin infrastructure under direct federal oversight. The new bank will initially provide digital asset custody services for Circle and its affiliates, with a pathway to serve institutional clients and potentially manage the USDC reserve in the future. The approval advances Circle from the conditional stage it reached alongside other crypto firms in late 2025. This move reflects a broader trend of integrating crypto infrastructure into the existing U.S. banking framework through national trust charters, which specialize in custody and fiduciary services rather than traditional lending.

ambcryptoHá 25m

Circle secures final OCC approval for national trust bank to strengthen USDC infrastructure

ambcryptoHá 25m

DeXe hits new ATH, targets $40 – THESE 2 indicators back bulls

DeXe (DEXE) surged to a new all-time high of $35.5, currently trading around $34 with a 17.9% daily gain. The rally is supported by strong market participation, with trading volume jumping 161% to $128 million. Buyers have dominated the market for eight consecutive days, showing aggressive spot accumulation. Additionally, open interest rose 18% to $160 million, indicating active new positions in derivatives. Key technical indicators like the RSI (76) and the ADX (with +DI above ADX and -DI) suggest strong buying pressure and a solid uptrend. If demand persists and the $30 support holds, DEXE could target the $40 level.

ambcryptoHá 1h

DeXe hits new ATH, targets $40 – THESE 2 indicators back bulls

ambcryptoHá 1h

Is Ethereum Truly a "World Computer"?

Title: Is Ethereum Really a "World Computer"? Ethereum, envisioned as a "world computer" by its founder Vitalik Buterin, aims to be a decentralized platform for global applications. However, a recent analysis by Four Pillars raises questions about whether it is more accurately a "Western computer," based on the geographical distribution of its validators. Currently, the United States dominates with 38.19% of all validators, followed by Germany at 13.04%. Combined, these two countries account for over half of the network. In contrast, Asian representation is minimal, with Singapore holding only 3.15%. The concentration is partly due to affordable cloud hosting services like Hetzner and OVH in Europe and North America, as well as the prevalence of residential validators in the U.S., where individuals run nodes via home internet connections. When examining professionally operated validators, the distribution becomes more balanced. The U.S. share drops to 25.81%, while Asian countries like Singapore (7.28%), Hong Kong (6.44%), Japan (6.38%), and South Korea (4.59%) collectively approach the U.S. level. This shift reflects strategic deployments by institutions to meet regulatory requirements and reduce latency for local users. However, regions like South America, the Middle East, and Africa remain underrepresented. Ethereum's peer-to-peer network mechanisms, such as gossipsub, disadvantage areas with low node density, creating a feedback loop where delayed message propagation reduces validator performance and rewards. This imbalance challenges Ethereum's promises of censorship resistance and global accessibility. Despite these issues, opportunities exist for growth in underrepresented regions. As demand for localized staking infrastructure rises, early entrants in areas like the Middle East could establish dominant positions by offering compliant, low-latency solutions. The evolving validator landscape highlights both the structural challenges and the potential for Ethereum to move closer to its "world computer" ideal.

Foresight NewsHá 2h

Is Ethereum Truly a "World Computer"?

Foresight NewsHá 2h

Trading

Spot
活动图片

Categorias popularesright-arrow

Etiquetas Popularesright-arrow