Isomorphic Labs and Google DeepMind outline joint bioresilience strategy

·4 views

Isomorphic Labs and Google DeepMind have jointly published their approach to bioresilience, detailing how they plan to address biological risks using AI-driven research. The post appears on DeepMind's official blog, signaling a coordinated effort between the two organizations. Bioresilience refers to building defenses and preparedness against biological threats, including pandemics and engineered pathogens. The collaboration reflects growing interest in applying advanced AI tools to global biosecurity challenges.

Read the full story at Hacker News

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

Discussion (0)

How to Convert PDFs to Images in the Browser Without Uploading Files

Developers can convert PDF files to PNG or JPG images entirely within the browser using Mozilla's pdf.js library, eliminating the need to upload sensitive documents to third-party servers. The library, which powers Firefox's built-in PDF viewer, renders each PDF page onto an HTML canvas element that can then be exported as an image. A key technical consideration is resolution: PDFs are defined in points at 72 DPI by default, so developers must apply a scale factor to achieve sharper output, with 150 DPI recommended for screens and 300 DPI for print-quality results. When exporting to JPG, a white background must be painted first since the format does not support transparency, unlike PNG. The client-side approach offers privacy, speed, and zero backend costs, though rendering large documents at high DPI can be memory-intensive on the user's device.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How CDN Configurations Can Block AI Crawlers Like GPTBot From Your Site

Websites that rank well on Google may still be invisible to AI platforms like ChatGPT due to hidden CDN-level blocks that standard robots.txt checks won't reveal. Crawlers such as GPTBot, ClaudeBot, and PerplexityBot can be silently blocked by CDN rules that operate independently of a site's robots.txt file. A developer guide published on DEV Community on July 19 outlines how to test whether these AI bots can access your content. The process covers detecting hidden crawler blocks across multiple bots, including Bingbot, and is described as achievable in under five minutes. Making content accessible to AI crawlers is increasingly important as platforms like ChatGPT become significant sources of traffic and visibility.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

27% of Sites Accidentally Block AI Crawlers at CDN Layer, Not robots.txt

An audit of several thousand websites found that roughly 27% block at least one major AI crawler, with most blocks happening unintentionally at the CDN or WAF layer rather than through robots.txt settings. Common bot-protection rules — such as blocking unknown user agents or challenging non-browser traffic — inadvertently stop crawlers like GPTBot, ClaudeBot, and PerplexityBot. This means a site's robots.txt can appear fully permissive while the server silently returns a 403 error to AI bots. Developers can test this by sending curl requests with AI crawler user agents and checking edge logs for 403 responses grouped by user agent. Experts note that a clean robots.txt is a necessary but insufficient condition — the CDN configuration is what ultimately determines whether AI engines can access a site's content.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Self-Destructing launchd Job to Auto-Migrate Claude Code Model Settings

A developer has shared a fully automated macOS launchd script designed to update the model field in Claude Code's settings.json on a specific date — July 7, 2026 — ahead of the claude-fable-5 model's end of life. The script backs up the existing settings file, rewrites the model value using jq, validates the resulting JSON, and sends a desktop notification upon completion. A key feature is self-destruction: after executing, the job unregisters itself via launchctl unload, preventing any future accidental re-runs. A date guard at the start of the script defends against launchd's catch-up behavior, which can trigger missed scheduled jobs on Mac reboot before the target date. The author describes the pattern as general-purpose for any one-time, date-specific configuration change that must happen exactly once without human intervention.

0 comments Read more at DEV Community