Valyu's Science MCP Server Gives AI Agents Access to Full Research Papers

·1 views

Most AI research agents rely only on paper abstracts, missing critical details buried in methods sections, figures, and supplementary data. Valyu is a search API designed for AI agents that retrieves full-text scientific literature, clinical trial records, patents, and financial filings with structured metadata and resolvable identifiers. The company offers a hosted MCP server accessible via a plain HTTP endpoint, requiring no local setup, and compatible with tools like Claude Desktop and Cursor. Developers can connect it using a Valyu API key, with free credits available upon sign-up, and control retrieval scope through dataset IDs or curated presets covering fields from biomedicine to chemistry. The platform indexes over 37 million open-access biomedical papers via PubMed alone, alongside arXiv, ClinicalTrials.gov, and more than 36 other datasets.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Same AI Model, Same Benchmark: Scores Range from 30% to 100% Based on Test Setup

Between July and August 2025, multiple research teams tested the same AI models on the ARC-AGI-3 public benchmark and reported scores ranging from 30.16% to 100%, without any change to the underlying model weights. ARC Prize officially verified Claude Opus 5 at 30.16% using a deliberately basic harness, while MIT and NVIDIA later reported 100% on the same model using more capable testing frameworks. OpenAI demonstrated that simply enabling retained reasoning and output compaction tripled GPT-5.6 Sol's score from 13.3% to 38.3%, revealing how much the surrounding infrastructure influences results. None of the perfect scores have been verified on the private benchmark set, and the researchers themselves acknowledge their results do not isolate the model's contribution from the harness. The growing gap between official and optimized harness scores raises serious concerns about what AI benchmark numbers actually measure and whether they can be meaningfully compared without full disclosure of testing conditions.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

AI Is Reshaping Developer Hiring, Not Eliminating the Role Entirely

Artificial intelligence tools are accelerating software development tasks such as code generation, debugging, and documentation, allowing smaller teams to match the output of larger ones. This productivity shift means companies may require fewer developers for routine work, altering hiring expectations across the tech industry. However, core engineering decisions—such as system architecture, security design, and failure handling—still require human judgment that AI cannot reliably replace. Developers who lack foundational knowledge risk accumulating technical debt by blindly accepting AI-generated code that may contain logic errors, security flaws, or performance issues. The most valuable skill is increasingly the ability to combine domain expertise with AI tools to solve complex problems efficiently and critically.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Senior Engineer Logs 54 Hours of AI-Assisted Dev Work to Map the Productivity J-Curve

A senior software engineer and tech lead publicly documented 19 AI-assisted development sessions totaling an estimated 54 hours of work, tracking time across implementation, specification, infrastructure, and debugging. The data revealed that only 20% of total effort went toward writing feature code, while 80% was consumed by infrastructure, specification, and debugging work. Infrastructure time as a share of total effort climbed steadily from 27% in early sessions to 71% in later ones before dropping sharply to 21% in the final assembly phase. This pattern — a prolonged productivity trough before an uptick — is what the author calls the 'Productivity J-Curve,' a concept attributed to frameworks developed by Dan Shapiro and AI strategist Nate B. Jones. The engineer frames the findings as a candid, real-time account of learning to work at higher levels of AI-assisted software development.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Laptops Are the New Security Perimeter as Credentials Pile Up Unmonitored

Enterprise security has shifted across three eras: from defending corporate networks, to managing digital identities, and now to securing the developer endpoint itself. Modern developer laptops routinely store cloud access keys, API tokens, SSH keys, and secrets cached by AI coding agents in config files and dotfiles. Existing controls leave critical gaps — firewalls never see these credentials, identity providers only govern login events, and endpoint detection tools monitor process behavior rather than plaintext secrets at rest. An attacker who gains access to such a device can exploit valid, authorized credentials without needing to breach any authentication system. Security experts argue the next phase of perimeter defense must begin with full visibility into credentials already residing on developer machines.

0 comments Read more at DEV Community