←Search results

Search “Agents”

By relevance · Total 6730 results Clear search

Saturday

  1. Lauren Tan

    How to trust the code produced by AI agents more

    Lauren Tan proposes using a four-layer inverted pyramid to increase trust in AI agent code: codebase improvements, static analysis, prose-style rules (such as AGENTS.md and skills), and human review, with investment concentrated in the topmost layer. She suggests handing this post directly to your own bot or agent, and provides steps for installing the tool, creating a new Project, and using commands such as /recall, /poteto-mode, /correct, and /architect to analyze historical commits and feedback.

    Lauren Tan proposes using a four-layer inverted pyramid to increase trust in AI agent code: codebase improvements, static analysis, prose-style rules (such as AGENTS.md and skills), and human review, with investment concentrated in the topmost layer. She suggests handing this post directly to your own bot or agent, and provides steps for installing the tool, creating a new Project, and using commands such as /recall, /poteto-mode, /correct, and /architect to analyze historical commits and feedba

    Matched fieldsSummary、Body

    #hot-signal#Tutorials / practice#Agents#English text is being prepared.
  2. Arena

    Arena tracks four signals for agents from six major AI labs

    Arena tracked four Agent Arena signals for the best agents from Anthropic, OpenAI, Google DeepMind, Kimi, xAI, and DeepSeek from late June to early October, finding that the largest gap between labs lies in whether users enjoy the experience of using them. On Praise vs Complaint, OpenAI and Anthropic scored +35 points above the leaderboard average, while Moonshot, xAI, and DeepSeek ranged from +5 to +13; on Confirmed Success, the average lab's best model rose from -4 to +14 points within three months, with Anthropic leading at +20 with Claude Opus 5.5 (Max). The Bash Recovery gap narrowed from 34 to 16 points, and Steerability from 19 to 11 points.

    Arena tracked four Agent Arena signals for the best agents from Anthropic, OpenAI, Google DeepMind, Kimi, xAI, and DeepSeek from late June to early October, finding that the largest gap between labs lies in whether users enjoy the experience of using them. On Praise vs Complaint, OpenAI and Anthropic scored +35 points above the leaderboard average, while Moonshot, xAI, and DeepSeek ranged from +5 to +13; on Confirmed Success, the average lab's best model rose from -4 to +14 points within three m

    Matched fieldsSummary、Body

    #editorial#English text is being prepared.#Agents#Evaluations / benchmarks
  3. OpenAI Developers

    How OpenAI Designs Failover for Model Providers

    For customers such as those in financial services, OpenAI explains the production routing and failover design for critical AI workloads between the direct OpenAI API (1P) and third-party cloud providers (3P). The direct API allows early access to preview models and features, and failover is enabled only when a request can be safely retried and the fallback route meets the requirements—for example, in a payment flow, a timed-out Responses API call for a read-only fraud assessment can be retried before the transaction deadline. In-progress sessions in the Agents API cannot be recovered by switching providers; multi-provider support is a platform capability, while failover is a workflow decision.

    (3P). The direct API allows early access to preview models and features, and failover is enabled only when a request can be safely retried and the fallback route meets the requirements—for example, in a payment flow, a timed-out Responses API call for a read-only fraud assessment can be retried before the transaction deadline. In-progress sessions in the Agents API cannot be recovered by switching providers; multi-provider support is a platform capability, while failover is a workflow decision.

    Matched fieldsSummary、Body

    #engineering#Tutorials / practice#English text is being prepared.#OpenAI

Friday

  1. heise onlineSelected

    Meta Releases 816B Open-Source Model Muse Spark Open, Partnering with Apple for Local Execution

    Meta has released the open-weight model Muse Spark Open, which uses a MoE architecture with 256 experts, activates about 42B parameters per token, has a context window of 1 million tokens, and supports text, image, video, and audio input but outputs text only. Meta has partnered with Apple to run the approximately 530 GByte four-bit quantized version on two interconnected Mac Studio M5 Ultra machines, distributing computation via the MLX framework and Thunderbolt 5, achieving about 25 output tokens/second at an 8192-token context. Artificial Analysis tests show that it scores 39 on the Intelligence Index, making it the strongest American open-weight model currently available, but it trails MiMo-V2.6-Pro's 46 points and GLM-5.3's 45 points, and ranks last in four Agent tests. The model weights are available on Hugging Face in two versions: 16-bit at approximately 1.6 TByte and four-bit at approximately 530 GByte, with API pricing at $1.25 per million input tokens and $4.25 per million output tokens.

    an 8192-token context. Artificial Analysis tests show that it scores 39 on the Intelligence Index, making it the strongest American open-weight model currently available, but it trails MiMo-V2.6-Pro's 46 points and GLM-5.3's 45 points, and ranks last in four Agent tests. The model weights are available on Hugging Face in two versions: 16-bit at approximately 1.6 TByte and four-bit at approximately 530 GByte, with API pricing at $1.25 per million input tokens and $4.25 per million output tokens.

    Matched fieldsSummary

    #hot-signal#Model releases#Open-source ecosystem#Agents
  3. Claude Blog

    How an Anthropic Japan BDR Uses a Team of Claude Agents to Handle Daily Work

    An Anthropic Japan business development representative used Claude Managed Agents (beta) to build an agent team called Nakama, handing lead triage, qualification, email drafting, Salesforce updates, and post-meeting follow-up to 6 agents that run on a schedule, completing most of the work before the start of each day. He first prototyped it in Claude using skills, scheduled tasks, and a shared folder; after migrating, skills became system prompts, the shared folder became memory storage, and changes are proposed uniformly in a Slack channel and synced by a coordinating agent roughly every 15 minutes.

    An Anthropic Japan business development representative used Claude Managed Agents (beta) to build an agent team called Nakama, handing lead triage, qualification, email drafting, Salesforce updates, and post-meeting follow-up to 6 agents that run on a schedule, completing most of the work before the start of each day. He first prototyped it in Claude using skills, scheduled tasks, and a shared folder; after migrating, skills became system prompts, the shared folder became memory storage, and cha

    Matched fieldsSummary、Body

    #lab#Tutorials / practice#Agents#English text is being prepared.

Thursday