AI Coding Live News

Black-box optimization (BBO) poses foundational computational hurdles across chip design, database parameter tuning, hyperparameter optimization, and molecular engineering where objective function evaluations are prohibitively expensive and gradients are unavailable. Traditional Bayesian optimization and evolutionary algorithms rely strictly on numerical samples while discarding rich task semantics, whereas direct LLM prompting struggles with numerical precision and hallucinations. Researchers from Nanjing University's National Key Laboratory for Novel Software Technology (LAMDA Group) introduce AgenticBBO-Bench (arXiv:2610.12183), the first unified cross-domain benchmark for evaluating LLM agents in black-box optimization. Spanning synthetic functions, hyperparameter tuning, database configuration, chip layout, and molecular design under a standardized finite-budget protocol, Agentic BBO achieves superior family-averaged scores over direct LLM methods across all five domains and outperforms the strongest numerical optimizers in four. Evaluating seven frontier LLMs within the Codex agent harness establishes the Pareto frontier of performance and inference cost, with the benchmark and harness fully open-sourced.

Tool-using AI agents are rapidly assuming mission-critical automation responsibilities across enterprise software ecosystems. However, standard agent benchmarks focus exclusively on nominal, forward task completion—conflating initial planning competency with operational resilience and fault recovery. In real-world enterprise environments where network drops, idempotency failures, and partial commits are commonplace, unguided LLM retries routinely cause devastating side effects such as duplicate payments or corrupted databases. Researchers introduce UndoBench (arXiv:2610.05622), a comprehensive evaluation suite spanning 36 enterprise workflows and 36 production fault scenarios across 8 enterprise verticals. Utilizing counterfactual paired trials alongside wire-level effect-history and state oracles, UndoBench decouples task competence from recovery capability. Experiments reveal a severe vulnerability gap: while nominal task competence achieves 83.54%, conditional recovery success rates (CRSR) collapse to 46.72%, with naive retry producing duplicated external mutations in 53.33% of trials. The code is available at GitHub (tradertanmay/undobench).

@hithere_devs and @AyanMn18 launched Client SQL (clientsql.com), a PostgreSQL control plane for Codex, Claude Code, and Cursor: per-person and per-agent scopes, agent grants cannot exceed the human owner, writes and schema changes need an independent human reviewer, plus audit trails and per-client revocation. The hosted product is still a pilot; the site advises starting with a disposable or staging database.

@DanKornas introduced feigeCode/navop, a native desktop workspace that keeps database workbenches, SSH/local terminals, SFTP, and remote desktop in one app. AI features cover SQL help, terminal assistance, and an Agent Hub; builders can attach Codex, Claude Code, and OpenCode over ACP and optionally expose host tools via MCP. The repository had about 1.4k GitHub stars at verification.

Databricks engineer Yuchen Jin says Databricks Unity Gateway is coming to developers on Neon (@neondatabase) and calls it the fastest AI Gateway for Kimi K3, noting Databricks inference is moving quickly. For teams building agents/apps on Neon that need low-latency Moonshot Kimi K3 access, this is a new managed gateway path.

LangChain and TypeSafe AI released Jev, a specialized system model designed to classify and guardrail dangerous tool calls. Jev intercepts raw shell commands, database drops, and cloud API triggers with deterministic type validation to prevent catastrophic AI agent errors.

JetBrains released Rider and ReSharper 2026.2.2 featuring a dedicated AI Agent Setup widget. Developers can visually discover, configure, and authenticate MCP-compliant external tools, databases, and agent skills directly within the IDE.

Perplexity says two engineers worked with hundreds of AI coding agents to build CobbleDB, a custom Rust database replacing DynamoDB reads in its search stack. The case shows coding agents contributing at real infrastructure scale while keeping production execution outside the agents’ permissions.

Keewano launched KeewanoDB alongside a $12M seed round. The database is designed for machine reasoning by preserving ordered event context for users, devices, and agents, so agents can read coherent history instead of reconstructing it from conventional tables.