A queryable long-term memory for AI coding agents using a two-tier system: an always-loaded index and an on-demand SQLite database, overcoming context window limits.

Simon Willison releases sqlite-utils 4.0rc2, largely developed with the help of Claude Fable. The article details the process from rc1 to rc2, including critical bug fixes like delete_where() not committing causing data loss. GPT-5.5 was also used for cross-review, uncovering two additional issues. The total estimated unsubsidized cost was $149.25.

Release of sqlite-utils 4.0rc2, a Python CLI utility and library for manipulating SQLite databases. Notably, this version was mostly written by AI Claude Fable, costing about $149.25. Announced by Simon Willison on July 5, 2026.

An open-source tool that bootstraps a personalized Obsidian vault with AI-guided onboarding and document ingestion, creating a structured knowledge base for AI agents.

Limboo is a local-first desktop workspace for AI software development, providing a connected environment for coding agents. It organizes work around sessions, integrating repositories, git, terminal, memory, and a security boundary, all in one place.

Alibaba's Damo Academy has introduced the first AI agent for discovering superconducting materials, called Elements Claw. It identified four new superconductors confirmed in lab experiments, screening millions of crystal structures in hours.

Nvidia is financing the neoclouds that buy its GPUs, renting back idle capacity and taking a share of their cloud revenue, gradually turning itself into something more than a hardware company.

This article explains that AI security is an extension of existing identity and authorization patterns, not a new discipline. It covers three AI use cases (RAG, tool use, agentic systems) through the lens of authentication and authorization, using a banking example. Key principles include deterministic identity layer, filtering in RAG, and chain of identity for agents.

Iwo Kadziela (with Codex assistance) created a credible ASCII world map using only 445 bytes of data, leveraging deflate compression and a clever JavaScript snippet using fetch() with data: URIs and DecompressionStream.

Armin Ronacher reports a curious problem: newer Claude models (Opus 4.8 and Sonnet 5) sometimes add invented fields to Pi's edit tool calls, causing rejection, while older models do not. He theorizes that Anthropic's reinforcement learning to optimize for Claude Code's built-in edit tool inadvertently degrades performance on third-party harnesses. This raises the question of whether frameworks like Pi should implement multiple edit tools to match model-specific optimizations.

A leaked video reveals Microsoft's exploration of a dedicated AI OS, codenamed Aion, built on Microsoft Edge and a lightweight Windows codebase called Win3. Centered around Copilot and web technologies, it lacks native Win32 app support but uses Cloud PC for desktop apps. Features include 'Spaces' for grouping apps/sites. While possibly just an experiment, its lessons may influence Windows 11.

Surge is an AI growth assistant for early-stage founders that creates weekly multi-platform content plans from a single startup URL, generating native posts for TikTok, LinkedIn, Reddit, and X.

This RFC proposes a real-time budget decision plane for AI agent runs, using atomic reservations and per-run ceilings to prevent uncontrolled spending, with machine-readable state for agent adaptation.

A study using Ramp spend data and Revelio Labs workforce data finds that companies investing heavily in artificial intelligence grow headcount by 10.2% over two years, with entry-level positions growing 12%. Gains are only for high-intensity adopters and subject to a learning curve. AI adoption is unevenly distributed.

Discusses the possibility of training AI to invert AES by treating it as a satisfiability problem, exploring both opaque and lowered circuit encodings, and considering complexity theory constraints like average-case hardness and phase transitions.

An ABC NEWS Verify investigation reveals that influencer Lily Jay has used AI-generated and manipulated videos to falsely claim charitable activities in Uganda, Gaza, and elsewhere, raising concerns about donor trust and NGO integrity.

A developer who doesn't know Rust used AI to build a PHP interpreter from scratch in Rust, which now passes 17.4% of PHP's official test suite and can render WordPress pages. The project relies on 22,000 tests as an objective oracle, avoiding AI self-evaluation bias.

The NHS will begin using AI on its app to triage patients and direct them to appropriate services, as part of a £10bn overhaul of NHS systems in England. The tool will help determine whether patients need a GP appointment or should be advised to visit a pharmacy or A&E. The update is expected to reach about 200,000 patients over the next year.

A video explores why AI can cost more than the human labor it replaces, examining real-world examples of hidden costs like hardware, maintenance, and energy.

Hanging up immediately on AI scam calls may not be the best approach. This article explores effective strategies, including staying calm, gathering evidence, and reporting the scam, to protect yourself and others.

ToothPaste is an open-source project that creates an encrypted BLE dongle for wirelessly transmitting keyboard and mouse commands to any USB-compatible device, without drivers or complex setup. It uses AES-256 encryption, ECDH key exchange, and an ATECC608B crypto coprocessor for security. It can be controlled via a web app using Web BLE, and includes AI agent integration.

Providence AI is an operator-led AI consultancy based in Wisconsin, USA, offering custom AI systems owned and run on your terms, with services from audit to ongoing stewardship.

The article draws an analogy between the common misconception that trees come from soil (when they actually come from air) and the tendency in AI safety to skip foundational concepts. Many AI safety students know specific details but cannot explain why AI is an existential risk, indicating a failure to internalize basics. The author argues that truly understanding the problem is what motivates sustained effort.

The federal health department has raised concerns about the use of AI scribes by doctors, as regulators consider safeguards. AI scribe tools have boomed in popularity in GP surgeries.

An open-source AI writing tool that detects 36 common patterns in AI-generated text, provides real-time rewriting, and includes a deterministic Python linter for CI integration.

A new study shows generative AI can design burgers that are tasty, sustainable, and nutritious by learning from large-scale recipe data.

Analysis of OpenRouter data shows US companies still account for most of the top 50 most-used AI models, but their presence has declined while China's has grown from 5 to 20 models. Few models come from other countries like Canada and France.

This article, part 1 of a series on AI memory, explains the difference between traditional computer storage and AI memory, covering context pools (RAG), semantic search and its time-blindness, vector embeddings, the distinction between training data and the vault, and how these mechanisms relate to hallucinations.

Kynto is an AI-powered recruitment assistant that goes beyond traditional Applicant Tracking Systems. It automates end-to-end recruitment tasks including job posting, candidate sourcing, evaluation, scheduling, and interview preparation. The platform learns a company's hiring preferences in minutes, integrates with existing tools, and aims to reduce bias and save time.

This guide explains why AI agents need a real inbox (not just send capability) and covers two architectures: bring your own agent or use MailKite's managed service. It addresses key needs like verification codes, async human-in-the-loop, conversation threads, and stable identity, and compares current agent-email platforms.