Context Mode keeps bulky tool output outside an agent’s prompt and restores relevant session state after compaction. Its MCP tools and hooks support 17 clients, though enforcement varies by platform.

OpenAI is publishing early internal data on how coding agents affect research workflows, but the supplied material names the measurements without reporting results.

Semantically equivalent instructions can make VLM reward models score identical robot behavior differently. Agent evaluators should test paraphrase invariance, not assume model scale or reasoning fixes it.

MiniMax argues that long-running, multimodal agents need enough context for repeated tool responses, while sparse attention makes that context more tractable.

Vercel AI Gateway now serves the medical-focused Ling 3.0 Flash Sante free through October 4. Use its free-only model ID to prevent requests from converting to paid usage afterward.

EDGE synthesizes multi-step tool-call training data by retaining only API links verified through live execution. Its 9B model nearly matched an untuned 27B sibling on Korean public-API tasks.

Vercel AI Gateway now exposes GPT-6 Astra as openai/gpt-6-astra for code and coding-agent integrations, targeting long-running tasks that combine software, browser, and data work.

LLMs' stated top decision factors only weakly tracked factors shown to affect outputs. For agent oversight, validate explanations with controlled interventions before using them for escalation.

Molecular benchmark accuracy can reflect retrieval of published values rather than prediction. Contamination audits should test digit-level recall and repeat runs across reasoning settings.

LLM decompilers can produce compilable C that passes shipped tests while changing behavior or hiding crashes. Validate reconstructed code against the original with shared fuzzed inputs.

A compact classifier distilled from LLM rationales handled product-pair decisions without inference-time LLM calls. Category adapters improved accuracy further while retaining large speed and cost gains.

ECC packages skills, hooks, memory, orchestration, and security controls for multiple coding-agent harnesses, but its breadth makes selective installation and verification essential.

Agent memory can break across model upgrades even when the store is unchanged. Test migrations by direction, rebuild embeddings completely, and retain raw histories for repair.

A 52,988-request audit found black-box LLM judges too unstable for preregistered gates. Measure repeatability first, and pilot the judge before fixing evaluation thresholds.

Knowledge-work agents need code-like infrastructure around tools: centralized context, action records, verification, enforced permissions, and preflight checks for irreversible work.

Cursor Cloud Agents can execute inside per-request Vercel microVMs, giving teams control over isolation, credentials, retries, and worker lifecycle without operating persistent VMs.

Playco reports that GPT-6 Astra halved manual fixes while producing three themed game prototypes from one grey-box base, suggesting less cleanup in model-driven iteration.

Readable chain-of-thought is not a reliable map of which steps caused a correct answer. Treat step-level LLM critiques as partial signals, especially on already-correct traces.

Shared agent memory needs user-scoped reads, human-approved writes, and per-request credentials. The talk offers a concrete security boundary for company-wide context and tools.

Legora says GPT-6 Astra reviewed 41 documents within minutes, caught every planted error, and improved its workflow result by nearly 40%, though the underlying measure is unspecified.

Controlled pre-training experiments suggest varied reformulations can teach facts more efficiently than repeating documents under the same token budget, though paraphrasing gains depend on batch size.

Cross-silo agents are primarily a context and privacy-boundary problem. Start with low-sensitivity data, explicit sharing policy, and approval at the moment information leaves a silo.

Marketing Skills packages shared product context and repeatable growth workflows for multiple coding agents, reducing setup across copy, CRO, SEO, analytics, pricing, and launches.

OpenAI introduces GPT-6 Astra with claimed advances in computer use, coding, cybersecurity, and science. The supplied material gives no benchmarks or implementation details to assess those gains.

On-policy distillation recovered most full-data gains from one query; diverse rollouts mattered more than dataset size, while slow student alignment remained the bottleneck.

Two Sigma runs remote agents under each employee’s real identity, adding trace propagation and controlled web access to preserve attribution and limit egress risk.

GPT-6 Astra is OpenAI's first model rated Critical for cybersecurity capability under its Preparedness Framework, a material consideration for security-sensitive agent access and controls.

SWE-Gate shows why green tests are an incomplete agent-eval signal: 221 of 644 functionally passing repairs still violated constraints derived from code review.

Zo packages a persistent Linux server, coding agents, files, integrations, and hosting into one personal cloud. The useful idea is collapsing build and deploy context into the agent’s workspace.

A causal framework separates deceptive-looking model output from the mechanism producing it, giving agent evaluators a stricter basis for claims about intent or agency.

Sentinel-RL keeps topology outside the LLM: a graph encoder and constrained policy choose SOC actions, while the model writes critic-gated narratives for human approval.

User feedback helps models repair targeted faults, but LLM judges often miss those repairs. Agent evals should retain human feedback and avoid treating model preference as ground truth.

Cursor cloud agents can now execute tools on machines and elastic pools inside your network while Cursor retains planning and inference. This unlocks internal access and custom hardware, not full self-hosting.

OpenAI has deprecated its standalone skills catalog. Builders should use the Plugins repository and current plugin guide for examples, including skill-only packaging.

Muse Spark 1.3 gives coding agents a 1M-token, multimodal model through Vercel, with a cheaper contributor tier that permits Meta to train on submitted inputs and outputs.

Language-based monitoring cannot fully reveal an agent’s internal computation, the paper argues. Secure agent runtimes still need isolation and data-flow controls independent of self-reporting.

GLM-5.3 is half-price through September 8 via a temporary DigitalOcean-only model ID. Keep the standard ID in durable agent configs if you need fallback after the offer.

A coding-specialized model paired post-training with an iterative generate-evaluate-refine loop to exceed the top IOI 2026 human score. The reusable idea is feedback-driven test-time search.

Gemini 3.8 Flash brings multimodal input, tool calling, web search, and default reasoning to coding agents through Vercel. Its temporary 50% discount runs through December 31.

A UE5M3 block-scaling recipe trained an 8B model in FP4 without Hadamard transforms or BF16 final layers, while reporting better losses and downstream estimates than the compared recipe.

Cliff trains on the location of a rollout’s first reasoning error, rewarding the valid prefix and penalizing the suffix. It offers finer supervision without a specialized reward model.

A telecom diagnosis pipeline structures telemetry before forcing a decision path and evidence-grounded explanation. The pattern is relevant to agents reasoning over heterogeneous operational data.

A label-free Dutch-book test finds internally inconsistent probabilities from language models, especially when prompts add logical complexity or irrelevant context.

SafeEvolve turns completed agent trajectories into reversible harness updates and policy training, linking runtime safeguards with learned multi-step behavior.

This talk proposes legally registered agent organizations, scoped JWT authority, and blockchain audit trails for open-internet commerce; it is an architecture proposal, not validation.

Generative UI needs more than a component catalog: agents require explicit layout rules, schemas, and design-system constraints to produce stable interfaces across repeated requests.

Paid agent tools turn a skill file into spending authority. Builders need budgets, transaction logs, and counterparty screening before letting agents purchase data or services unattended.

x402 servers can perform work before payment settlement, leaving a double-spend window. Builders should settle first or accept explicit counterparty risk until stronger schemes mature.

Agent authorization should scale with stakes and reversibility: logs may cover coding changes, while open-ended payments need scoped, time-bound, independently verifiable mandates.

AWS is separating agent payment policy from model execution: AgentCore handles wallets and limits, while WAF meters bot access. The useful pattern is deterministic spend control at the edge.

A wallet-equipped agent crossed paywalls and completed paid email and phone actions under a spending cap. The engineering lesson is to enforce budgets in the wallet, not in prompts.

For fuzzy requests, an agent should identify the missing constraint with the most decision value, elicit it in the right modality, then choose a response format suited to the task.

Agent checkout needs more than payment tools: enforce user-set limits outside the agent, expose structured merchant data, honor cancellation, and log every decision.