iCreatAI News
Back to main site
Skip to content
#Tutorials
Oct 29Thu
  1. Hacker News AI58

    Our AI Agent Executed the Same Successful SQL Query 254 Times

    Funnel compared the open-weight model GLM-5.2 against Claude Opus 4.8 in a randomised A/B test on its internal data agent platform Reef, finding GLM averaged 7.8 seconds per model call versus 18.1 seconds for Opus but made 66% more tool calls (19 vs 11) and took 14% longer end-to-end. In two separate runs GLM executed the same successful SQL query 113 and 254 times, in one case repeatedly querying a tool that only exposed a 20-row preview. The author attributes much of the problem to the agent harness—nonexistent Athena tables in saved instructions, ambiguous tool contracts and missing safeguards against repetitive calls—and recommends explicit tool contracts, bounded agent behaviour and whole-system benchmarking.

  2. ElevenLabs Blog23

    Conversational AI in banking: Use cases, benefits, examples

    This guide explains how banks use conversational AI—NLP-based chat and voice agents that understand natural language, pull account context, and escalate to humans when needed—for 24/7 support, onboarding, disputes, and fraud alerts. It cites CFPB findings on early bank chatbots, McKinsey figures on reduced handling time and call volume, and ElevenLabs customer examples from Customers Bank and Klarna, while detailing security standards such as SOC 2 Type II, ISO 27001, PCI DSS, and Zero Retention Mode.

  3. AWS Machine Learning Blog23

    Accelerating software delivery with agentic QA automation using Amazon Nova Act – Part 3

    This third post in the series describes two QA Studio capabilities for scaling agentic QA across an organization: applications, which group use cases and test suites under an owner and report health via pass rate, a flaky-adjusted 0–100 score, trend charts, and failure triage built on Amazon CloudWatch custom metrics; and team-based user management, which assigns five cumulative roles (Admin, Team Lead, Editor, Executor, Viewer) through Amazon Cognito groups, with cross-team sharing capped by the recipient's role. The backend enforces access through a Lambda authorizer resolving Cognito groups into scopes plus a resource check on application ownership or sharing, and the reference solution is deployed with the AWS CDK.

  4. AWS Machine Learning Blog36

    Distributed training with Ray Train DLCs and EFA on EKS

    This AWS post shows how to fine-tune Qwen2.5-1.5B with PyTorch FSDP across eight NVIDIA L4 GPUs on two EFA-enabled g6.12xlarge instances on Amazon EKS, using the published Ray Train DLC (public.ecr.aws/deep-learning-containers/ray:train-ml-cuda-v1.1) without building any image. KubeRay creates the Ray head and worker pods from one RayCluster manifest, while hardware, transport and per-job Python packages are configured from outside the image via resource requests, environment variables and Ray runtime_env. The sample repository includes idempotent deploy scripts that locate g6.12xlarge capacity, provision the cluster and node group, install the GPU and EFA device plugins and KubeRay, and submit the job from inside the head pod, with the script failing if NCCL did not use EFA.

  5. AWS Machine Learning Blog42

    Building an Agentic AI Platform on Amazon Bedrock AgentCore

    This AWS post describes an enterprise Agentic AI Platform blueprint on Amazon Bedrock AgentCore that separates a centrally governed Platform control plane from repeatable, team-owned Workstream cells, with separate inference and tool gateways and a federated operating model. It details multi-account topology, service control policies, a CI/CD pipeline with deployed-runtime evaluation gates and adversarial denial tests, and dependency-ordered retirement, using the eu-west-1 (Ireland) Region. The companion GitHub blueprint lets teams deploy a representative Workstream cell and validate its security and lifecycle contracts end to end.

  6. Google Developers Blog58

    Hill-Climbing Agent Quality with Agents CLI Evaluation

    A walkthrough of the Agents CLI eval loop using the LangChain template in Agents CLI v1.8.0, covering a baseline run (mean 4.0), a system-prompt regression that dropped the mean to 2.75, per-case diffs via eval compare, and a gemini-3.8-flash to gemini-2.5-flash swap whose mean drop traced to a single off-topic case. It also shows scheduling eval runs with cron or Cloud Scheduler, a script that exits non-zero when scores fall past a threshold, and eval optimize for automated prompt hill-climbing on ADK projects.

  7. Hacker News AI58

    C# and NATS.io for agentic AI orchestration

    This post argues that production agentic AI systems are distributed systems, and introduces the open source Synadia Agent Protocol for NATS, a contract (not a framework or runtime) that lets agents of any kind be discovered, prompted, and monitored on a single NATS fabric. The protocol defines three responsibilities—discovery via NATS micro service ping/info on the "agents" service, conversation via typed JSON chunks streamed to a reply inbox with a mandatory first ack and a zero-byte terminator, and liveness via heartbeats on agents.hb.{agent}.{owner}.{name}—along with a subject hierarchy, request envelope, and error reporting conventions. Reference SDKs exist in TypeScript and Python, with chunked attachments, JetStream at-least-once delivery, stronger identity, and audit traces deferred to future extensions.

  8. Pydantic AI Articles74

    Code mode: make your agents 2x faster and 2x cheaper on long tasks

    A benchmark of 5,276 graded runs across eight models on five small Pydantic AI apps found that code mode, where the model writes a program that calls tools inside a sandbox, cut cost by 58% and wall time by 51% on long tasks while raising success by 16 points. Gains grew with task length: at 200 items code mode was 47% to 75% cheaper and 42% to 67% faster, whereas at 5 items it cost 21% to 38% more on three of five tasks. The author attributes the win not to fewer round trips, since models already batch calls, but to less content per request, and runs the model's code in Monty, a Rust Python interpreter with no host access by default and 30-second, 256 MiB limits.

    Why it matters: Offers a detailed, methodology-heavy benchmark with per-model cost, latency and correctness numbers, plus an honest list of where code mode costs more.

  9. Baseten Blog55

    The fastest, lowest-cost batch transcription with Qwen3-ASR

    Baseten optimized Qwen3-ASR 1.7B, an Apache 2.0 open-weight speech recognition model supporting over 30 languages and 22 Chinese dialects with a 5.4% average WER on the Open ASR Leaderboard, for batch transcription on a single RTX PRO 6000. The optimized stack transcribes 7x more audio per second than out-of-the-box vLLM at high concurrency, reaching 2,645x real-time factor on 5-minute clips and 3,128x on hour-long files, with median latency for 5-minute clips at 64 concurrent requests falling from 50 seconds to 7.3 seconds. Baseten reports a cost of about $0.0013 per audio hour, 8x lower than the same model on vanilla vLLM and roughly 277x less than popular closed-weight providers.

  10. Hacker News AI42

    Rent the model, own the harness: how to build a model-agnostic AI agent

    Werk describes building a model-agnostic agent harness for mobile teams, arguing the model understands and proposes while the harness decides what it reads, touches and records. It outlines four practices: calling models through an in-house interface, having agents request a level of work rather than a model, using a consistent input format, and placing stable content first for caching. It notes that agent SDKs handle the model-tool loop but not team-specific concerns like shared records, permissions and approvals tied to what a person saw.

  11. Hacker News AI55

    Progressive Determinism for AI Tasks

    The post describes a workflow in which an AI agent first explores a web application dynamically with Playwright to complete a scenario, then saves a deterministic Playwright test (e.g., tests/add-task.spec.ts) that is preferred on later runs. If the saved test fails after a UI change, the agent falls back to discovery mode, diagnoses the failure, and proposes a minimal repair that is kept only after the original assertion passes. The author notes that Playwright has built-in planner, generator and healer agents supporting this pattern, which they have not yet tried.

  12. Hacker News AI60

    Who Is the Agent Acting As? Why AI Agents Become Confused Deputies

    The article explains that AI agents given a single broad service account act with the platform's authority rather than the user's, so a prompt injection buried in a support ticket can produce a fully authorized write to another tenant's data. It recommends authorizing every tool call as the intersection of user, agent, and tenant permissions, decided in a gateway outside the model that fails closed and logs each decision.

  13. OpenRouter Announcements58

    Hit Your Claude Code Usage Limit? How to Keep Coding

    Claude plans have a five-hour limit and a weekly limit, and when you hit one you can keep coding by running Claude Code on an OpenRouter API key, which has no such windows and bills per token instead. The setup uses three environment variables (ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN, and an empty ANTHROPIC_API_KEY) placed in a shell function so plain `claude` still uses your plan, and you switch back by typing `claude` when the plan resets. The article reports median session costs of $0.305 on Opus 5.5, $0.155 on Sonnet 5.5, and $0.007 on Haiku 5.5, and warns that Claude Code's printed cost is unreliable because it prices OpenRouter models with Anthropic's table.

  14. OpenRouter Announcements38

    Claude 529 Overloaded: What It Means and How to Fix It

    A 529 overloaded_error means Anthropic's API is temporarily out of capacity for the requested model across all users, unlike a 429, which reflects your own usage limits. The article recommends retrying with exponential backoff and jitter, then routing the same model to another provider via OpenRouter, which serves Claude Sonnet, Opus and Haiku 5.5 through five providers each, and optionally listing a fallback model with `models`. For Claude Code, it suggests switching models with `/model` or running Claude Code on OpenRouter via Ori Harness or three environment variables.

  15. Allen Institute for AI Blog38

    How researchers are adapting olmOCR 2 for historical and archaeological archives

    Ai2 describes how researchers adapted its open olmOCR 2 model: historian Maurice Casey built ArtificeOCR to run olmOCR 2 locally and make his Tropy photo library of archival sources searchable, while TU Wien and Austrian Academy of Sciences researchers fine-tuned olmOCR 2 with LoRA on 57 annotated Carnuntum pottery records, lowering transcription error on 450 test records from 15.58% to 1.11% through their CENTURIA benchmark.

Oct 28Wed
  1. Hacker News AI28

    How to Build a Slow Product in the Fast AI Era

    The developer of Mole, a Mac and Windows cleanup tool, describes spending two months on unscalable work: installing and uninstalling 850 Mac apps one by one to build cleanup rules, testing more than 30 AI tools in real use, and redrawing dozens of icons over three rounds. Mole went from 1.13 to 1.18, added AI tool and System Data cleanup, shipped a Windows version on October 14, and passed 72.9K GitHub stars. The author argues that as AI makes code cheaper, the details learned only by doing the work yourself increasingly set products apart.

  2. Hacker News AI36

    The half-life of AI knowledge

    The article argues that AI knowledge decays at different rates across layers: tools and products last months to a couple of years, techniques a few years, architectures five to ten years, and core ideas decades. It cites examples such as OpenAI's Agent Builder shutdown, the expiry of chain-of-thought prompt phrasing, and the long lineage of gradient descent (1847) and backpropagation (1986). It recommends learning threshold concepts so new releases read as variations on ideas already understood.

  4. Hacker News AI43

    Scaling AI in Legal: Building Uber's Redlining Agent

    Uber describes building its first-generation Legal Redlining Agent (LRA), a Microsoft Word add-in that reviews counterparty redlines, detects intent, recommends policy-aligned responses, generates comments, and flags risks for lawyer review. Through four iterations—RAG, lawyer feedback loops, tone modulation, and agentic modifications plus a rules database—Uber reports over 20% reduction in average contract review time and 91% accuracy in AI-generated decisions. The post details the architecture (React/TypeScript add-in, Python back end, LangGraph workflow, OpenSearch vector stores, Redis session memory) and time-weighted feedback sampling to mitigate concept drift.

  6. Hacker News AI40

    The AI cost of JavaScript frameworks

    A benchmark had coding agents build the same small PokéAPI app five times in each of twelve JavaScript frameworks/meta-frameworks, then measured cost, time, tokens, lines of code, and Lighthouse performance and accessibility. Hono+htmx and Astro were cheapest (about $2.67–$2.76 per run) while Angular, Gracile+Lit and SvelteKit were most expensive, and JavaScript bytes ranged from 0 KiB (Astro) to 142 KiB (Next.js). The author presents the results as evidence that framework choices can be made on data rather than vibes, while noting the experiment is stochastic and limited to one content-heavy app type.

  7. Hacker News AI50

    Decompiling Games with AI: Virtueel Zoetermeer

    The author describes using Claude to decompile Virtueel Zoetermeer, a Dutch municipality-commissioned SimCity-style serious game from the early 2000s, after an attempt to decompile Age of Mythology: Gold Edition stalled at around 60%. Claude reconstructed equivalent Pascal source code, which is published on GitHub, and the game is now playable online at virtueelzoetermeer.bou.ke.

  8. Hacker News AI38

    Building My Own Tiny AI Weather Model

    The author builds a small graph neural network weather model inspired by Graphcast, using a 5 GB subset of ERA5 data (North Atlantic/Europe, 2003–2016 training, 2017–2019 evaluation) and an encode-process-decode architecture with an icosahedral mesh. The model has about 200k weights and trains in 45 minutes on a five-year-old gaming PC, outperforming the persistence baseline for all nine variables at lead times up to +72 h, though its RMSE is roughly 5x worse than Graphcast at 24 h and 10x worse at 72 h.

  9. Hacker News AI36

    Getting good at AI takes time and practice

    The post argues that becoming skilled with AI requires durable, long-term abilities rather than disposable prompt tricks, including supplying the right context, learning what models can and can't do, recognizing characteristic AI mistakes when reviewing code, and quickly skimming verbose output. It contends that new models improve mostly on tasks older models already handled well, so these skills remain useful, and that software's human-machine "centaur age" may last decades. The author concludes that deliberately avoiding AI likely leaves people behind, while going fully into AI is also a mistake since LLMs reward expertise best gained by hand.

  10. Hacker News AI18

    Building an AI Predictive Stock Algorithm

    The post links to a blog article at getfintab.com titled "Building an AI Predictive Stock Algorithm," with discussion on Hacker News (item 50269713). The supplied material contains only the title, URLs and submission metadata, with no article body to summarize.

  11. Hacker News AI60

    I ran each AI coding-agent sandbox's own tests against the others

    An open lab that turns sandbox claims (no key reads, no network, no git hooks) into runnable contracts and publishes results across sandboxes on the same tests. Corpus v1.1 (2026-10-26) holds 196 contracts harvested from the public test suites of sandbox-runtime, nono, Codex, Deno, bubblewrap and OpenHands, each re-expressed in a neutral format and credited. First results measure sandbox-runtime 0.0.79, nono 0.80.0, Codex CLI 0.162.1 and landrun on Linux kernel 6.8, with later runs adding TartarOS and masked-credential contracts.

  12. Hacker News AI58

    Writing AI agents in YAML instead of Python

    This article walks through Docker's open-source docker-agent tool, which defines an AI agent's model, instructions, and toolsets in a single YAML file and runs it with one command. Using a Polars documentation Q&A agent, it demonstrates a built-in RAG toolset with local SQLite vector storage, per-run model overrides, prompt files, flavors for variants, multiple agents sharing tools, JSON-based evals, and pushing agents to a registry. It notes that code-first frameworks like LangGraph suit cases requiring custom branches or retries.

  13. Hacker News AI40

    Wikipedia: Signs of AI Writing

    A Wikipedia field guide catalogues writing and formatting conventions typical of AI chatbots such as ChatGPT, with real examples drawn from Wikipedia articles, drafts and comments, to help editors detect undisclosed AI-generated content. It stresses that the listed patterns are only potential signs rather than proof, that AI detection tools and human judgment both have high error rates, and that the deeper policy problems behind the surface tells should be addressed rather than merely the symptoms.

  14. ElevenLabs Blog38

    How to build an AI voice assistant: Brand-aligned voice AI

    This guide explains how to build an AI voice assistant by connecting speech to text, an LLM, and text to speech into one real-time pipeline, then adding a knowledge base, business tools, system instructions, and fallback rules. It covers architecture choices between raw APIs and the managed ElevenAgents platform, latency targets below 500 ms with voice activity detection, caller verification, human handoff, brand voice customization, and pre-launch testing for latency, interruptions, accuracy, and safety. Post-launch metrics such as containment, escalation rate, word error rate, hallucination rate, and CSAT are recommended for continued improvement.

  15. Browser Use Blog38

    Why Playwright Scripts Break When the Site Changes

    The article explains that Playwright scripts break on redesigns because they locate elements by page structure (CSS classes, DOM position, XPath, generated IDs) rather than by what users see, and recommends role, label, text, or owned test-ID locators with auto-waiting instead of fixed timeouts. It catalogs causes and fixes for flakiness, including strict-mode violations, copy changes, A/B variants, iframes and shadow DOM, expired sessions, and overlays handled via page.addLocatorHandler() since v1.42. For sites you don't control, it lays out options from manual fixes to Playwright's healer agent and agent fallbacks, with a Python example attaching Browser Use to Chrome over CDP.

  16. Browser Use Blog38

    How to Run Browser Automation at Scale in the Cloud (2026)

    This guide describes a queue-and-worker architecture for cloud browser automation, where each job gets its own isolated remote browser driven over CDP with Playwright or Puppeteer, with concurrency capped at the plan limit and browsers always stopped when jobs end. It compares pricing, concurrency limits and session caps across seven providers (Browser Use, Browserbase, Steel, Hyperbrowser, Browserless, Kernel, Anchor), works through a cost example of 1,000 jobs a day, and covers rate limits, retries with backoff, and self-hosting versus managed trade-offs.

  17. Browser Use Blog30

    Headless Browser Blocked by Cloudflare? What Works (2026)

    Cloudflare blocks stock headless Chrome because it signals automation (navigator.webdriver is true, Playwright's headless shell sends HeadlessChrome in its user agent), often runs from cloud server IPs, and starts each run without cookies. The article recommends getting recognized via an official API or Web Bot Auth verification, or, on sites you're allowed to automate, using a real Chromium with residential IPs and a persistent profile, which in practice means a managed stealth browser. It cites Stealth Bench V2 (September 25, 2026) results where headless Chrome with no proxy got through 3% of visits versus 90.5% for Browser Use Cloud, noting the site set favors the vendor.

  18. Hacker News AI38

    AI search explained: indexing, retrieval and reranking

    This post walks through three retrieval approaches on a synthetic ten-note dataset from eight tools: text search using an analyzer and inverted index, embedding search using an embedding model and vector database, and embedding search with a reranker that reorders a shortlist by relevance. It shows text search missing paraphrases, embedding search surfacing semantically similar notes but ranking a refund how-to above the actual reasons, and reranking putting the relevant notes on top. It concludes with guidance on when to use each approach and notes that production systems often combine text and embedding search via hybrid search with reciprocal rank fusion before reranking.

  19. Claude Blog38

    Manage costs and demonstrate ROI for Claude Code on Google Cloud

    This post outlines a control, track, and prove framework for managing Claude Code spending on Google Cloud, using native tools such as per-project billing, budget alerts, spend caps, BigQuery exports, quotas, and the self-hosted Claude apps gateway for per-user and per-group spend limits and model allowlists. It describes tracking via Claude Code's OpenTelemetry metrics and events exported to Cloud Monitoring or BigQuery, and proving ROI through usage, satisfaction, and impact metrics such as PR cycle time and review turnaround. It cites Delivery Hero, Shopify, and Fifth Dimension as examples of measured impact.

  20. Hacker News AI38

    Securing sensitive data in the AI era

    This article explains shadow AI—unsanctioned use of AI tools at work—and how it appears in development teams, noting that client code and personal data carry extra contractual and GDPR obligations. It offers practical mitigations: approved business-plan tools, a shared model gateway, sending less data, pseudonymization, IDE file exclusions, local models, and agreeing AI rules with clients at kick-off. It also covers incident response and stresses rewarding reporting rather than punishing it.

  21. Hacker News AI38

    How to match sizes and colors across stores so AI agents can buy the right one

    Channel3 explains how it builds a product catalog for AI agents by storing products rather than web pages, using per-category rules to decide which variations count as the same product, reading option values instead of labels, aligning each store's size and color buttons to the brand's naming, and mapping values onto fixed per-category lists so search can filter exactly. The article walks through adidas Gazelle Indoor and Nike Air Force 1 examples, noting that over 90% of apparel values, over 95% of shoe values and over 75% of furniture values map onto these lists. It argues structured listings let an agent query every listing at once and get consistent answers, unlike web search built for people.

  22. Hacker News AI22

    Vibe coded a point-and-click game to learn about the Drupal AI Ecosystem

    A developer vibe coded "AI Mac Packwala and the Curse of the Stale Cache," a browser-based point-and-click adventure in which players escape an island by learning the Drupal AI ecosystem, including providers, automators, agents, Tool API tools and MCP. The game has about 60 puzzles over three acts, is built in vanilla TypeScript with Vite and drawn entirely in code on a 320 × 200 canvas, and is also available as an Electron desktop app for Windows, macOS and Linux. The source code and a walkthrough are available.

  23. Databricks Blog33

    Unlocking Investment Insights Without Migrating Your Data: Lakehouse Federation

    Databricks Lakehouse Federation lets users create federated connections to external databases such as SQL Server, PostgreSQL, MySQL, and Snowflake and query them directly from Databricks with standard SQL, keeping data in place. The post walks through a wealth management example that joins on-premises portfolio holdings with FactSet data subscribed from the Databricks Marketplace in a single query, then powers a dashboard and a Genie natural-language agent without pipelines, materialized views, or persisted objects. It notes the same pattern applies to other industries, such as healthcare, manufacturing, and retail.

  24. Google Developers Blog34

    Building Conversational Commerce with Multi-Agent Swarms on GCP: A Working Prototype

    The post presents HappyDelivery, a working prototype of the previously described three-layer conversational commerce architecture built with LangGraph, Gemini, FastAPI, and Google Cloud. It details an omni-channel ingress service that verifies platform-specific signatures and normalizes payloads before queuing them to Pub/Sub, a Redis session and cart layer, and a LangGraph agent swarm with supervisor, profile, inventory, and transaction specialists using deterministic pricing tools. Live benchmarks and a three-turn simulated shopping flow, plus a five-test automated suite, are reported as verification.

  25. Pragmatic Engineer58

    AI SRE: what is it, how could it change on-call – and does it work?

    This Pragmatic Engineer article examines the rise of "AI SRE" incident tooling, driven by rising deploy frequency and pager volume, drawing on interviews with Rootly, incident.io and ResolveAI. It describes how these agents mostly speed up incident investigation inside Slack using context layers, knowledge graphs and memory, while whether they should make code changes remains debated. It also covers accuracy tactics like adversarial verifier agents and backtesting, latency/cost tradeoffs, in-house builds at OpenAI, Stripe, Google and Monzo, and the limits of fully autonomous remediation.