A security skill router gives coding agents scoped, repeatable playbooks for reverse engineering and pentesting instead of ad hoc tool selection. Its case workflow also preserves evidence and findings.

MCP Apps let servers return sandboxed, interactive UI instead of text alone while the host retains control of prompts and tool calls—a useful pattern for agent workflows needing charts or controls.

Vercel AI Gateway now exposes Qwen 3.8 Max to coding agents, adding one model endpoint for long-context text and vision work with gateway routing, budgets, and usage tracking.

MCP Tasks aim to make long-running tool calls durable across disconnects and crashes. The proposed V2 removes session-heavy state, but polling scale and implementation complexity remain unresolved.

Nick Heiner argues that leaderboard gains can diverge from useful agent behavior through contamination, weak verifiers, reward hacking, and test conditions that users cannot inspect.

Cybersecurity agents need deterministic exploit oracles and tasks that reward finding distinct vulnerabilities, not self-reported wins or one crash. A capability ladder makes partial progress measurable.

Human task duration is a weak proxy for agent difficulty when agents use different shortcuts and tools. Long-horizon evals need environment complexity, queryable trajectories, and judges that inspect final state.

RLHF can make agents persuasive assistants without making them dependable autonomous decision-makers. Builders should separate human-pleasing interaction from calibrated automation and keep stakes bounded.

Agent Reach installs, checks, and routes web-access tools for coding agents, reducing the setup work behind research across social platforms, video sites, GitHub, RSS, and the open web.

TokTier targets a hidden agent-serving bottleneck: repeatedly tokenizing long transcripts after small tool results. Its stateful repair cut median vLLM time to first token by 16–34% in reported tests.

AI Gateway can now enforce spend caps across a team, project, or API key, giving agent workloads layered cost controls instead of relying on per-key limits alone.

Training-data curation can improve model quality and inference efficiency without simply adding compute. The practical work is decontamination, deduplication, balancing, task matching, and selective synthesis.

AgentHPOBench tests whether agents can learn from experiment history, not merely produce code. Its results expose weaknesses in sustained refinement and log diagnosis across sequential ML runs.

DeepSeek V4 Flash’s updated weights replace the preview behind the existing model ID, raising its reported Terminal-Bench score from 56.9 to 82.7 without code changes.

Training on a production harness could adapt an agent to real workflows, but non-replayable interactions break common RL assumptions and make feedback harder to turn into updates.

OVI suggests interactive expert feedback can let a smaller learner imitate values without representing the expert's full policy. The result matters to agent distillation, but relies on specific assumptions.

An open-source memory hub turns agent conversations, workflows, docs, and code into governed assets that can be reused across sessions and roles, reducing repeated project setup.

Chat SDK’s Teams adapter adds reactions, user-only ephemeral messages, custom token supply, and routing fixes useful for building safer agent approval and progress flows.

Post-training gains depend heavily on task selection, rollout quality, and environment design. For many enterprise agents, curated SFT may deliver most of the value before costly RL.

FriendBench shows why aggregate accuracy can hide behavioral bias: top multimodal models matched human panels overall but favored the “stranger” answer and gained less from video.

AI Gateway’s dedicated logs expose per-request cost, tokens, latency, routing, and provider fallbacks, making agent failures and spend anomalies easier to trace.

Long-running agents strain context, reward design, and GPU scheduling at once. Compaction and value models can help, but each introduces another optimization trade-off.

A multi-policy PEFT method groups and sequences compatible tasks into separate QLoRA paths, reducing interference without adding trainable capacity; it reached 44.78 on TRACE.

Vercel MCP now serves the 2026-07-28 and 2025 protocols from one endpoint, letting newer clients adopt stateless requests and revised authorization without breaking older setups.

Reasonix is a configurable terminal coding agent built to preserve DeepSeek prefix-cache reuse across long sessions, with optional planner/executor models and subprocess-based plugins.

Repo-only tasks omit deployments, incidents, traffic, and distributed failures. Training infrastructure agents may require multi-node environments with real resources, despite much higher setup cost.

AI Gateway has raised Laguna S 2.1 capacity tenfold for both paid and free model IDs, reducing throughput constraints for high-volume or long-running coding agents.

Base-model data is shifting from broad web imitation toward code, reasoning, and agent-task priors. The unresolved choice is how early to introduce synthetic and instruction-shaped data.

AI Gateway now exposes one beta fast-mode option across models, letting coding agents request lower latency while retaining standard-speed fallback when no fast tier exists.

Biology agents need evaluators that verify analysis of large experimental datasets, not recall. LatchBio found human review essential because valid scientific paths can defeat brittle graders.

For subjective agent output, replace vague requests for quality with decomposed brand constraints, then reserve human preference data for style and creativity that resist deterministic checks.

Coding benchmark scores can mislead when prompts leak implementation details or tests reject valid solutions. Inspect tasks and verifiers before using a leaderboard to choose an agent.

Cursor’s cloud-agent adoption grew after it treated the dev environment as agent infrastructure: Linux parity, one discoverable CLI, end-to-end testing, and automated repair.

OpenAI says GPT-5.6 Luna and Terra now cost less, which may change model-routing choices for agent workflows. The supplied material gives no prices or workload comparisons.

Two API settings—reasoning retention and compaction—reportedly tripled GPT-5.6’s ARC-AGI-3 score. Agent evals should treat runtime configuration as part of the tested system.

OpenAI positions GPT-5.6 as delivering more useful output per dollar across inference and agent workflows. The supplied material has no metrics for judging routing or migration decisions.

OpenAI reports that scientists are using coding agents to modernize scientific software and accelerate work in genomics, though the supplied report summary offers no methods or results.

Gemini Managed Agents now add model selection, sandbox hooks, token caps, schedules, and free-tier access, covering several controls needed for unattended coding work.

Cursor Start gives developers in India a ₹649 monthly tier with Grok 4.5, Composer, cloud agents, mobile control, and UPI payments, positioned between Free and Pro.