One Useful Thing
·
1 month ago
Author Ethan Mollick announces a new book called Co-Existence, written partly with AI assistance, about working with AI systems that are sometimes better and sometimes worse than humans, following his earlier bestseller Co-Intelligence. Recent developments show AI coding agents leading to 17 times more code written and Anthropic reporting that AI now writes 80% of its code with each developer shipping 8x more output. The shift from human-centered co-intelligence with chatbots to autonomous AI agents fundamentally changes how humans will interact with AI across many fields beyond just software development.
The Algorithmic Bridge
·
1 month ago
Anthropic has filed confidentially for an IPO with a near-$50B annualized revenue run rate and approaching profitability, while OpenAI trails with lower revenue and no profitability in sight despite both companies burning substantial cash. Both companies are moving toward public markets while their business models remain unprofitable at scale, with recent examples showing enterprises like Microsoft, Uber, and Starbucks cutting AI spending due to unsustainable costs and reliability issues. If these companies' IPOs precede a market realization that AI promises exceed actual returns, early investors will transfer accumulated risk to pension funds and the broader economy rather than realizing gains from genuine technological success.
Google Research
·
1 month ago
Google researchers developed PHRM, a system that uses a smartphone's front-facing camera to passively measure heart rate during everyday use by analyzing facial video clips with deep learning. The system achieved mean absolute percentage error less than 10% for heart rate measurement and mean absolute error less than 5 beats per minute for daily resting heart rate compared to wearables, with equal accuracy across all skin tones using data from over 350,000 video clips from nearly 700 diverse participants. This enables smartphone-based heart health monitoring to reach the 5 billion smartphone users globally, particularly benefiting people in low-resource settings who lack access to wearable devices.
Hugging Face Blog
·
1 month ago
NVIDIA released Nemotron 3.5 Content Safety, a 4B-parameter model that evaluates text, images, and responses simultaneously to detect policy violations that emerge only from their interaction. The model achieves 85% average accuracy across multilingual and multimodal safety benchmarks while maintaining low latency on 8GB+ VRAM GPUs. Enterprises can now enforce custom domain-specific policies at inference time and receive auditable reasoning traces explaining each safety decision, rather than relying on fixed universal taxonomies.
Deep Learning Weekly
·
1 month ago
Anthropic released Claude Opus 4.8 with improved coding and agentic performance, OpenAI expanded Codex with role-specific plugins for non-developers, and MiniMax launched M3 with frontier coding, multimodality, and 1M-token context. Claude Opus 4.8 maintains the same pricing as 4.7 while adding benchmark gains and reducing code flaws. These releases change the landscape by making advanced AI capabilities more accessible across different roles and increasing open-weight model competitiveness through larger context windows and broader functionality.
Ben's Bites
·
1 month ago
A newsletter discusses progress on building an agents manual and reflects on how creating tools enables further tool-building, followed by a roundup of recent AI releases including new open models, agent products, and agent-focused tooling across multiple domains.
IBM Research
·
1 month ago
IBM released Granite Libraries, a set of adapter tools that break down large language models into modular, task-specific components similar to software building blocks, allowing developers to customize AI behavior without retraining entire models. The Granite 4.1 8B model with requirement-check adapters achieved 84% balanced accuracy on the IFEval instruction-following benchmark, compared to 51% for the base model with prompting alone. This modularity enables multiple teams to incrementally add capabilities, reduce inference costs, and make AI systems more predictable and maintainable for enterprise deployment.
OpenAI Blog
·
1 month ago
Endava is integrating AI agents and ChatGPT Enterprise into its software delivery processes to automate workflows and accelerate development cycles. The company is deploying these tools across enterprise operations to build what it describes as an AI-native organizational culture. This shift aims to reduce manual tasks in software development and enable teams to focus on higher-level strategic work.
OpenAI Blog
·
1 month ago
OpenAI has added a memory system to ChatGPT that retains user preferences and context across multiple conversations. The feature allows ChatGPT to recall details about individual users without requiring them to re-explain preferences in each session. Users now receive more personalized responses based on accumulated interaction history rather than starting from scratch with every conversation.
Hugging Face Blog
·
1 month ago
Hugging Face redesigned its hf command-line interface to work efficiently with coding agents like Claude Code and Codex, which now drive a significant portion of Hub traffic. Testing showed that agents using the hf CLI consumed 1.3 to 1.8 times fewer tokens than those using curl or the Python SDK directly on complex multi-step tasks, with the CLI achieving 93-94% success rates versus 84-92% without it. The optimized CLI auto-detects agent usage, outputs dense TSV format instead of human-readable tables, and provides chainable next-step hints to reduce the number of commands agents need to execute.
OpenAI Blog
·
1 month ago
The U.S. government has outlined a strategy to integrate artificial intelligence into biological defense systems and pandemic preparedness. The plan focuses on using AI to accelerate pathogen detection, vaccine development, and epidemiological modeling across federal agencies. Implementation could reduce response times to biological threats and improve disease surveillance capabilities nationwide.
