iCreatAI News
Back to main site
Skip to content
#Tutorials
Nov 12Thu
  1. OpenRouter Announcements58

    How to Stop an AI Agent Stuck in an Infinite Loop

    OpenRouter describes four guards against runaway agent loops, tested on six models with its Agent SDK (@openrouter/agent 0.11.0): doom-loop detection, a loopKey that normalizes tool arguments, stopWhen caps on steps and cost, and a credit limit on the API key. With doomLoop: true, all six models were blocked on their third identical tool call, and across 18 runs detection cut tool calls from 314 to 64 and spend from $0.29 to $0.05. The article notes that loops vary between runs and that only the key-level limit caps total spend when agent code is buggy or caps are missing.

  2. AWS Machine Learning Blog46

    Detecting and recovering from silent data corruption on Amazon SageMaker HyperPod

    This post describes adding three checks to a PyTorch DDP training loop to detect silent data corruption (SDC), where hardware such as a GPU returns a wrong but plausible number without raising an error: a non-finite gradient check, a ring replay audit that recomputes another node's micro-batch and compares loss bits and gradient fingerprints, and a weight fingerprint comparison after the optimizer step. When a check fires, the suspect rank is mapped to its node via the Kubernetes node name, the node is cordoned, and SageMaker HyperPod replaces it, with training resuming from a checkpoint saved before the suspect window. The checks were tested with software-injected faults on a HyperPod cluster with 4 ml.p5.48xlarge instances (32 NVIDIA H100 GPUs) and repeated on three instances, with full code in an accompanying GitHub repository.

  3. AWS Machine Learning Blog34

    Semantic diffing of formal policies with anti-unification

    The post describes an open source engine that computes semantic diffs of formal policies by anti-unification over an e-graph, using declared algebraic laws and domain rewrites so that shared structure appears once and each genuine disagreement is kept as a Variants node. Worked examples cover SMT-LIB-style rules, Cedar and Dogwood authorization policies, and OpenShell and CEL encodings, showing how presentational differences are absorbed while real bugs such as a weakened conjunction or a dropped guard remain visible. The engine builds from source with a pinned Rust toolchain and reads a single S-expression file.

  4. AWS Machine Learning Blog38

    OpenAI Codex on Amazon Bedrock

    This AWS post explains how developers can run OpenAI Codex on the workstation while routing model inference through Amazon Bedrock using supported OpenAI models such as GPT-5.6 Sol, Terra, and Luna. It compares three access paths—Native AWS Access with IAM Identity Center, AgentCore Gateway, and an LLM Gateway—and recommends starting with a small, least-privilege pilot before expanding. It also covers validation steps, identity-isolation testing, OpenTelemetry and CloudWatch observability, cost allocation, plugins, hooks, MCP servers, phased rollout, and cleanup.

  5. Hacker News AI42

    Composing data pipelines in an AI-native workflow

    Fivetran describes its CLI as a terminal-native interface that lets engineers and AI coding harnesses (Claude Code, Codex, Gemini CLI, dbt Wizard) create connections, configure schemas, trigger syncs, and build custom connectors, with MCP servers supplying context from task trackers and observability stacks. The post walks through a session that lands Stripe data in Snowflake and ships an fct_mrr dbt model, then covers iterating on changed requirements and failed syncs, and offers guidance on scoped API keys, confirmation prompts, and prompt injection.

  6. Databricks Blog28

    Detecting credential-driven attacks on Databricks

    Databricks describes how credential-driven attacks unfold on its platform in six non-linear stages—initial access with valid credentials, environment mapping, credential harvesting, execution via UDFs and jobs, persistence, and data collection with cleanup—and what to look for in verbose audit logs at each stage. The post explains how to investigate these attacks with Genie and the Detection App, how to respond once a compromise is confirmed, and which preventive controls such as context-based ingress, least privilege, and credential rotation limit what a stolen credential can do.

  7. ElevenLabs Blog22

    Voice cloning for content creators: Scale your content workflows with voice AI

    This guide explains how voice cloning turns a recorded voice sample into a digital replica that can narrate new scripts for videos, podcasts, courses, and social clips, covering the capture-modeling-generation process and the difference between Instant and Professional Voice Cloning. It lists criteria for choosing a tool—voice fidelity, delivery control, language coverage, consent safeguards, export options, and commercial licensing—and outlines workflows for podcasters, YouTube, short-form content, courses, and audio articles. It also describes ElevenLabs' consent confirmation, voice verification, restricted voices, and traceability safeguards, plus which plan tiers include cloning and commercial rights.

  8. Browser Use Blog34

    How to Use Custom Proxies with Browser Use

    This tutorial shows how to connect a third-party (bring-your-own) proxy to Browser Use Cloud by supplying the provider's host, port, protocol, and credentials when creating an agent run or cloud browser, then verifying the browser's exit IP and country via ipwho.is. It also demonstrates using the same credentials with Playwright both locally and against a Browser Use cloud browser over CDP, and compares managed proxy pricing ($5/GB) with BYO (provider rate plus $0.20/GB Browser Use egress). The guide includes setup steps, code samples, a troubleshooting table, and production tips such as preserving cookies alongside sticky sessions and stopping unused browsers.

  9. Pydantic AI Articles34

    What we found testing OpAMP clients against the spec

    Pydantic describes how its Logfire Fleet Management OpAMP clients for Rust, Go, Node.js, Python, Java and .NET share one Rust protocol core, run as WebAssembly or a native library, and are checked against roughly 300 normative spec statements, TLA+ models, Kani and Verus proofs, black-box conformance and fault testing. Running the same scenarios against upstream OpenTelemetry clients surfaced bugs in the Collector's OpAMP extension and Supervisor and in the Java, Python and .NET clients, including instance UID handling, reconnect and status reporting, capability declarations and startup panics, plus one critical security vulnerability that the core is not affected by.

  10. Hacker News AI20

    An AI-assisted RP2040 audio board that worked on first spin in 5 days

    Beyondex is a custom hardware mod that brings Valve Index audio to the Bigscreen Beyond VR headset, combining an RP2040-based DAC, battery management, and mechanical integration. After two failed designer attempts, the team delivered a production-ready design in five days, replacing the IP5310 power approach with TI's BQ25185 and fitting the board to a 24 mm by 50 mm envelope. From initial contact to shipping 200+ production units took weeks, not months.

  11. Hacker News AI40

    Apple II Agent: An AI Coding Environment for the Apple II

    The author revived an open source Apple II Agent project in which an Apple IIe client connects over an Uthernet II Ethernet card, via a Mac proxy with Claude API access, to Anthropic's Claude Opus model so AI can write software through a chat on the machine itself. After automated testing in emulators on the proxy, the agent deploys new BASIC or assembly programs directly into real Apple II memory over the Ethernet link, and the author has made mouse-supported solitaire games and a Sim City clone. The code is on GitHub, the proxy runs on a Unix-like machine, and the Apple II end currently requires an Uthernet II card though it could be adapted to other network interfaces.

  12. AWS Machine Learning Blog40

    How DXC Technology built a multi-model AI substrate on Amazon Bedrock

    DXC Technology built DXC OASIS, a harness for mission-critical IT operations, on a governed multi-model substrate in Amazon Bedrock that provides access to Anthropic Claude, Amazon Nova, Meta Llama, and open-weight models through one managed interface, with Bedrock Guardrails, AWS PrivateLink, and serverless scaling as system-level controls. DXC separates the substrate from its own orchestration, governance, and assurance layer, and reports that since general availability in March 2026, DXC OASIS has reduced resolution time by 95% on average across 56 measured use cases. The post describes the architecture, model routing, and authority tiers, and states that DXC is evaluating Amazon Bedrock AgentCore alongside its current orchestration layer.

  13. AWS Machine Learning Blog34

    Automating AWS Agent Registry updates with push-based architecture

    This post describes an event-driven pipeline using AWS CloudTrail, Amazon EventBridge, and AWS Lambda that detects Amazon Bedrock AgentCore UpdateAgentRuntime events and syncs MCP server tool schemas into AWS Agent Registry records, supporting both single- and multi-account setups. The Lambda extracts the runtime ARN, obtains an OAuth bearer token via AgentCore Identity, calls the MCP server's initialize and tools/list methods, then updates the matching APPROVED registry record only when normalized tool definitions differ. Updated records revert to Draft status and require curator approval before becoming discoverable, and the post notes limitations such as no automatic registration of new servers and no retry on cold starts.

  14. AWS Machine Learning Blog42

    Accelerating generative AI delivery with reusable assets

    AWS's Generative AI Innovation Center, now part of AWS Frontier AI Engineering & Services, offers a catalog of 23 production-ready open source reusable assets such as RAG pipelines, document processors, and agent scaffolding. The post cites DoorDash and Lyft using the Contact Center GenAI Agent and Ricoh using the GenAI Intelligent Document Processing accelerator, reporting results including a 50x increase in testing capacity and a 7-fold throughput increase. It also walks through deploying the Contact Center GenAI Agent and lists the full asset catalog on GitHub and AWS Labs.

  15. Hacker News AI60

    Prompt Injection, Esq.: pwning AI agent guardrails with lawyerspeak

    A staff member at an AI agent security company bypassed the read-only guardrail on a ChatGPT workspace agent connected to a Notion hiring tracker by posting an image of a fake law-firm letter granting edit permission, which the agent adopted as new instructions. The agent then asked clarifying questions per the fake letter's rules, and only failed to add a new entry because the Notion connection lacked a create tool, though it did have update access. The author argues permissions live in the attached tool list rather than the prompt, and recommends scoping tokens or requiring human approval for writes.

  16. Comfy Blog58

    Doubling LLM token rates in ComfyUI

    ComfyUI rewrote the hot path of local text generation for Qwen3.x LLMs, letting the same checkpoints on the same GPU generate tokens up to twice as fast with no change to model or output quality; on an RTX 5090, Qwen3.5 9B went from 106 to 219 tokens per second. The changes include a re-laid-out GEMV kernel, fused prologue operations, new attention and DeltaNet decode kernels, and an adaptive-tree MTP speculative decoding algorithm, all available in ComfyUI v0.45.0 with the Text Generate node's mtp option left on auto.

  17. Amazon Science33

    Using mathematics to validate network configurations

    AWS describes using its automated-reasoning tool Tiros to statically validate network segmentation in Cloud WAN by converting configurations from route tables, attachments, and policy documents into logical propositions and querying which attachments can reach which with which destination IPs. The system checks that no path exists between segments the policy declares off limits, succeeding if actual reachability matches expected segmentation and alerting an operator if they differ, with analyses possible as often as hourly. The post illustrates this with a production/development segmentation example across us-east-1 and us-west-2 and notes the approach extends to tools like IAM Access Analyzer and VPC Reachability Analyzer.

  18. Meta Engineering40

    Intelligent Delivery: Making Every Unit of Compute Serve Better Ads

    Meta describes Intelligent Delivery, an ML-driven system that allocates compute across its ads serving stack toward a system-wide optimum, contributing roughly 10% incremental ranking performance on top of model and system advancements from 2017 through the first half of 2026. It combines a shared Resource Core Model (6B sparse and 20M dense parameters) that predicts each request's outcome, compute consumption, and latency impact, with a hierarchy of allocation engines spanning region, hardware SKU, and host-level decisions. Agentic AI now helps propose and test optimizations, contributing to about 20% of Intelligent Delivery optimizations in the first half of 2026 and about 80% of engineering effort on the most automated mechanism, with engineers approving every change.

  19. AWS Machine Learning Blog42

    Accelerating product development with AI-driven lifecycle management (AI-PDLC) in AWS Quick

    This AWS blog post describes AI-PDLC, a framework of six AWS Quick agents (an Orchestrator plus Discovery, Prioritize, Working Backwards, Spec Generator, and Designer) that guide product teams through the Discover, Decide, Design, and Prototype phases, with human approval required at every gate. The agents pull context from enterprise connectors such as Slack, Jira, Salesforce, SharePoint, ServiceNow, and Outlook, store artifacts in shared Quick Spaces, and produce a prototype_spec.md that can be handed to coding agents like Kiro or Claude Code via the Agent Client Protocol (ACP). The post includes step-by-step setup instructions, Deep Discovery and Fast Track pacing modes, and three scenarios: new product development, feature additions, and bug fixes.

  20. Databricks Blog38

    Air gapped, not data dark: how Bright Machines gives customers live factory dashboards without opening the plant

    Bright Machines captures PostgreSQL changes at the factory edge with Debezium, writes them as Apache Iceberg, pushes them out over a single auditable outbound path, and reads them in place in Databricks via Unity Catalog federation, with zero inbound ports to the plant. A row committed on the line is queryable in Databricks in one to two minutes, dashboards refresh every ten minutes, and a heartbeat check detects a broken pipeline in under thirty minutes. The first customer-visible dashboard went live about two months after kickoff, replacing nine loosely coupled services with one governed platform plus one edge path.

  21. Hacker News AI58

    Top 10 things AI gets wrong in web development

    Google released Modern Web Guidance 1.0, a set of skills that embeds web platform expertise, best practices and browser compatibility data into coding agents. Its evaluation suite found unguided coding agents pass only about 50% of checks on average, versus about 90% with Modern Web Guidance installed. The post lists ten areas where models frequently err, including focus and inert management, reduced-motion support, form validation and autofill, swipe gestures, dialogs and popovers, dark mode, main-thread responsiveness, layout shift, DOM XSS defenses, and passkey sign-in.

  22. Hacker News AI58

    Asynchronous, Long-Running AI Agents with Google ADK

    This article argues that most real AI agents run for minutes to days and should be treated as asynchronous systems rather than chat sessions, and walks through building async agents with Google's Agent Development Kit. It covers the Runner, SessionService (in-memory vs. database), Events, FunctionCalls and RequestInput, then shows a fire-and-forget pattern using HTTP 202 responses and webhooks, plus human-in-the-loop pausing and resuming via interrupt IDs and ResumabilityConfig. A second part is planned to cover production deployment with Cloud Run, Cloud Tasks and Cloud Scheduler.

Nov 11Wed
  1. Hacker News AI58

    The biggest AI cost reduction measure today is a cloud provider discount, lol

    AI spend has moved from experiment to budget line in about two years, with 98% of FinOps teams now managing some of it, and the article argues that committed-spend discounts are today's biggest cut because they lower token price without reducing token volume. It details three engineering levers that do reduce volume: prompt caching (cutting the input side of the bill by about 85%), batch APIs at half price on Anthropic, Bedrock and OpenAI, and smaller models for routine work, noting Claude Haiku costs about a fortieth of Opus per token. It then frames self-hosted LLM versus API as a per-workload choice among rent, route or own, with self-hosting paying off only at high, steady GPU utilization.

  2. Hugging Face Blog38

    Token Pooling-Aware Training for Multi-Vector Embedding Models

    This post explains token pooling for multi-vector embedding models, which merges similar document token vectors via Ward hierarchical clustering to cut storage, and introduces pooling-aware training in Sentence Transformers that applies the retrieval loss to both original and pooled representations. At pool_factor=4, the pooling-aware model retains 97.8% of its unpooled NanoBEIR nDCG@10 versus 94.6% for a normally trained model, with a 4x reduction in document vectors. It includes a reproducible training example, optional kernel acceleration, and evaluation results across pool factors, fixed caps, and calibrated adaptive thresholds.

  3. Hacker News AI38

    Modernizing NetHack's C engine with .NET MAUI, Avalonia and a 31-tool AI advisor

    Sound Mind Games (Hyvän mielen pelit ry), a Finnish nonprofit, describes how GnollHack wraps NetHack's roughly 500,000-line C99 engine as a native library behind a shared C# client, migrated from Xamarin.Forms and SkiaSharp to .NET MAUI by April 2025 and extended to Linux, the Steam Deck and macOS in 2026 via Avalonia.Controls.Maui. The game's opt-in Gnoll Overseer advisor embeds a web app that reads live game snapshots from the C engine and uses 31 tools (20 server-side, 11 on-device) to look up facts, and GnollBench grades models with the production prompt and tools using a blind two-provider judge panel, a geometric-mean score, a 25-point critical-error cap and a 50-point honesty floor. The post also covers the October 2026 nine-model benchmark results and the team's agentic workflow, in which coding agents write most C# and TypeScript code while the C engine, art, music and voiceovers remain human-made.

  4. vLLM Blog58

    Encoder Disaggregation in vLLM: Scaling Multimodal Serving with EPD

    vLLM details its Encoder/Prefill/Decode (EPD) disaggregation work, including metadata-only requests that let downstream instances reuse remote embeddings without repeating media preprocessing, a dedicated encoder-only Model Runner V2 path, and remote embedding transfer via NIXL and Mooncake. It also covers cross-encoder output reuse through Mooncake Store and orchestration support in vLLM Router, SMG, and llm-d, with core capabilities in vLLM v0.32.0 and DeepSeek V4.1 Flash metadata-only support on main via #62802. A DeepSeek V4.1 Flash benchmark on GB200 found EPD delivered 6.6–9.5% higher per-GPU throughput and 44–47% lower P90 TTFT than PD at concurrency 64–128, but lower per-GPU throughput at concurrency 16.

  5. Hacker News AI34

    We Run a Bootstrapped AI Startup on Almost Zero Operating Costs

    A former co-founder and CTO of a bootstrapped AI ed-tech startup describes how the team covered most product infrastructure through startup programs, credits and free tiers rather than paying for each SaaS service, including Sarvam LLM credits, Deepgram and ElevenLabs speech credits, Cloudflare, PostHog, Sentry, Neon and others. He details the mock-interview STT/TTS loop, concurrency limits across providers, and fallback routing, estimating roughly 235 concurrent interviews on paper with 20% headroom across three STT providers. He concludes that credits are not an architecture, since they expire and limits change, so integrations were kept swappable with fallbacks for critical parts.

  6. Latent Space28

    AI Engineer Code Summit: OpenAI on New Ways of Working With Codex

    At the AI Engineer Code Summit, OpenAI Developer Experience engineers Dominik Kundel and Charlie Guo presented a workshop on new ways of working with Codex, covering the revamped Codex Cloud, voice as a control plane for managing coding work, and personal agents called dots. Kundel acknowledged the cloud transition has been awkward and said OpenAI wants handoffs between local and cloud tasks to be easier, while Guo described dots as useful as a safety net, a delegation sink, and a ubiquitous presence.

  7. Hacker News AI42

    The Unexpected AI Stack: C# + .NET

    This post argues that C# and .NET are underrated foundations for agentic software development, citing build- and run-time type checks, Roslyn analyzers and source generators, EF Core schema checks, and a mature toolchain as guardrails for agent-written code. It highlights Aspire as a programmable runtime orchestration layer with isolated networking for worktrees and a queryable OpenTelemetry sink, and CSharpRepl as a way for agents to connect to a running application and wrap or replace functions at runtime. The piece is Part 1 of a five-part series and links the full repo at https://github.com/zeeq-ai/zeeq-tmpl.

  8. Hacker News AI20

    Controlling your AI agents from a game handheld

    The author describes using a Retroid Pocket 5 Android gaming handheld, paired with a portable OLED screen, power bank and foldable keyboard, to control personal AI agents on a mini PC during business trips. Software includes Termius, Tailscale, pi/opencode, herdr for multiplexing, and a web-based VS Code server for editing files and viewing diffs. The setup has so far been used in hotels and airports, and the author notes most Android terminal apps also support Mosh.

  9. Hacker News AI38

    AI proved your code correct. What did it actually prove?

    Keller AI explains how formal verification works, using a sorting example and Euclid's equilateral-triangle construction to show how proofs combine definitions, assumptions and prior results, then applies the approach to a document-download authorization handler. The article stresses that a checked proof only establishes a result within a mathematical model, so its value depends on whether the model's assumptions match reality and whether the specification captures the intended requirement, citing a revocation-versus-session example and timing side channels. It also reports an unreproduced synthetic experiment in which the best of seven automated reviewer setups approved 1 of 88 sealed rule-breaking changes.

  10. OpenRouter Announcements67

    MCP vs CLI vs Skills: What Each Costs in Context

    The article measures input-token costs for giving an agent GitHub access via the GitHub MCP server, the gh CLI, or a gh skill across Claude Haiku 5.5, Claude Sonnet 5.5, GPT-6 Luna, and Gemini 3.8 Flash; on Sonnet 5.5 the CLI used 1,421 tokens, the skill 3,315, and the MCP server 59,417 with its 46 default tools. It reports that the GitHub MCP server's 46 default tools add 19,436 tokens to every request on Sonnet 5.5, and that a single tool result (GitHub's release object) cost about 20,400 tokens. It also describes the beta Tool Search server tool, which cut the up-front cost of 46 tools to under 600 tokens and halved whole-task cost on Sonnet 5.5 while raising it on Haiku 5.5.

    Why it matters: It offers concrete, model-specific token and cost measurements for choosing between MCP, CLI, and skills, plus practical mitigations like scoping tools and trimming results.

  11. OpenRouter Announcements58

    LLM Quantization: What fp8 and fp4 Do to Quality and Price

    OpenRouter tested Qwen3.8 27B across six endpoints at fp16, fp8 and fp4 on 240 exact-answer tasks and found no detectable accuracy difference, with all endpoints scoring between 98.3% and 99.2%. Across 1,455 endpoints, fp8 was the most common labeled precision (23.6%) while 59.5% reported none, and lower precision was usually but not always cheaper. The post explains fp8 and fp4 formats and shows how to pin precision with provider.quantizations, noting that a narrow filter can leave a single endpoint exposed to rate limits.

  12. OpenRouter Announcements58

    LLM API Rate Limits: How to Fix OpenAI and Claude 429 Errors

    A guide to diagnosing 429 Too Many Requests errors from OpenAI, Anthropic and OpenRouter by reading rate-limit headers and limit_source, with tested fixes: retrying at the server-provided reset time, routing overflow to a fallback model, and letting another provider serve the same model. In the author's runs, default OpenAI SDK retries left 23 of 60 requests failed, while reset-aware retries completed all 60 in 89–92 seconds (Python) and 106–218 seconds (TypeScript), and a models fallback finished all 60 in 13 seconds for $0.000066. It also covers provider failover, which turned runs with up to 32 failures per 150 requests into 150 successes, plus habits for avoiding limits.

  13. AWS Machine Learning Blog40

    Build a daily dependency watcher with LangGraph and Amazon Bedrock AgentCore web search

    The Web Search API in Amazon Bedrock AgentCore is now available, returning current web results with title, URL, publication date and query-relevant extract, signed with existing AWS credentials and served from an Amazon-operated index. The post walks through building a LangGraph dependency watch that researches npm, PyPI and container base images in parallel, keeping the date window, domain blocklist and search budget in code and checking each finding against its cited sources. In the test run, three branches made 12 searches and finished in 36 seconds versus 99 seconds sequentially, and the watch can be hosted on AgentCore runtime or offered through AgentCore Gateway.

  14. ElevenLabs Blog12

    WAV vs MP3: How to choose the right audio file format

    WAV is an uncompressed, lossless format best for recording, editing and archiving, while MP3 is compressed and lossy, producing much smaller files suited to sharing and streaming. One minute of CD-quality stereo audio is about 10.6 MB as WAV, versus about 0.96 MB as a 128 kbps MP3 and about 2.4 MB as a 320 kbps MP3. Converting MP3 to WAV makes the file larger but cannot restore the data lost during compression.

  15. AWS Machine Learning Blog40

    The agent evaluation playbook: Built-in, DeepEval, and AutoEval on Amazon Bedrock AgentCore Evaluations

    This AWS post explains how to evaluate production agents on Amazon Bedrock AgentCore by starting from business outcomes, breaking them into observable behaviors, and mapping each to a trace- or session-level evaluator. It covers the built-in AWS evaluators, managed third-party metrics from DeepEval and AutoEval, custom and multi-dimension LLM-as-a-judge evaluators, and suggested evaluator stacks for customer support, research, and domain-content agents. It also describes three evaluation modes—on-demand, online, and batch—that share the same evaluator IDs, plus the ListEvaluators API for retrieving the full catalog.

  16. Pydantic AI Articles38

    Turn down OpenTelemetry noise without a redeploy

    Pydantic Logfire's Fleet Management lets you connect a service and its host Collector via OpAMP and change log level, attribute redaction and trace sampling while both keep running. The post walks through a FastAPI app with fleet=True, publishing SDK policies to drop debug logs, mask *token* attributes and sample 10% of traces, plus installing a host Collector and adding nginx logs. It also covers early-access rollouts, Kubernetes and existing Collectors, and troubleshooting.

  17. Pydantic AI Articles60

    Code mode is not a sandbox

    This post distinguishes code mode, where the model writes a short Python program calling your tools as functions via a single run_code tool, from a sandbox, which is an isolated environment where code runs. It describes three places a sandbox boundary can go (the whole agent, only the tools, or only the model-written code) and how Pydantic AI workspaces let you move that boundary without rewriting tools. It also covers Monty, a Rust-written Python interpreter that runs run_code next to the agent with no filesystem, network, or environment variables.

    Why it matters: Useful for teams deciding where to place isolation boundaries when combining code mode with sandboxed tool execution.

  18. Hacker News AI58

    An AI agent said the code was verified. The evidence said otherwise

    A blog post argues that when an AI agent implements, tests, reviews and reports on its own code, its claim of "verified" is untrustworthy, and proposes binding evidence (review, tests, human approval) to a specific Git tree hash so that changing the tree invalidates the evidence. It introduces Uncle (github.com/unclehq/uncle), an open-source tool that runs checks itself, records exit codes and tree identities, and distinguishes "verified," "signed" and "human looked at it" rather than blurring them. The author notes the tool does not establish correctness, is young, and is partly defeated by a hostile agent with unsandboxed shell access.