标签：#AI

Everyone can generate a website now. Type a prompt, get a decent page — that part is a commodity. The question nobody's answering is what happens on day 2 : the leads start arriving, a line of copy needs a tweak, someone asks for a section you forgot. That's when a website stops being a design project and becomes a thing you have to run — and where most tools hand you yet another dashboard to log into and dread. Sitelas makes a different bet. Because a Sitelas site lives inside Claude through an MCP connector, the same chat that built the site also runs it . You don't open an admin panel to see who filled out your form, write back, or change the page. You just ask. Here's what "running your site from a chat" actually looks like. First, the 30-second why Claude connects to outside tools through MCP connectors — you already use the ones for Gmail, Calendar, and Drive. Sitelas has one too. Add it once (in claude.ai: Customize → Connectors → Add custom connector , and paste https://sitelas.com/api/mcp ), and Claude can do things with your site, not just talk about it: publish it, read its submissions, restyle it, add a section. Your site becomes an automation endpoint sitting next to your other connectors — the thing a Webflow or Squarespace site can't be. New here? Start with How to Build a Website From a Claude Chat . "Did anyone fill out my form today?" That single question is the whole idea. You ask; Claude reads your site's submissions, surfaces the new lead — Maya, a bakery owner — and drafts a warm reply in your voice. One message, no tabs. It works because every form on a Sitelas site captures submissions to your inbox automatically — no integration required. You can open that inbox in the dashboard any time: …but running your site from a chat means you rarely need to. Claude reads those same submissions straight from your site, so "who wrote in today, and what do they want?" is answered in the thread you're already in — not in a panel you have to remember to ch

My CLAUDE.md has one line near the bottom that I wrote months ago and mostly forgot about until I started actually paying attention to what it does: ## Important Note after your work done codex will review what you done. Terse, no punctuation, clearly typed in a hurry. But it's a real instruction that fires on every session in this repo: I finish a change, and a second model reviews it before I consider the work done. I added it half as an experiment. A few months in, it's changed how I work more than almost anything else in the setup, and not in the way I expected. I thought it would catch bugs. Mostly it doesn't, not directly. What it actually does is force a triage decision on every single piece of feedback, and getting that triage wrong is where all the pain lives. The three buckets Early on I treated every review comment the same way: read it, do it. That lasted about a week before I was silently making changes I didn't agree with because a second AI suggested them, and separately burning a stupid amount of time re-litigating comments that were just wrong or out of scope. What actually works is sorting every comment into one of three buckets before touching code: Fix it, no discussion. The comment is unambiguous, low-risk, and doesn't touch anything architecturally significant. Just do it and move on. Ask first. The comment is ambiguous, or it touches something that would require a real judgment call, or the "fix" would be a bigger refactor than the comment implies. Stop and get a human decision before acting. Skip silently. The comment is a duplicate of something already handled, or genuinely doesn't apply. Don't reply just to say "not doing this," don't leave a comment thread as evidence of having read it. Silence is the correct response to a non-issue. The failure mode I kept falling into before I had these buckets explicitly was collapsing 2 into 1: treating "ambiguous" as "just pick an interpretation and go." That's the actual source of review fatigue, not

Introduction “When I use a word,” Humpty Dumpty said in rather a scornful tone, “it means just what I choose it to mean—neither more nor less.” “The question is,” said Alice, “whether you can make words mean so many different things.” “The question is,” said Humpty Dumpty, “which is to be master—that's all.” — Lewis Carroll, Through the Looking-Glass Humpty Dumpty believes that words can mean whatever we choose them to mean. Alice asks an interesting question. Can they? Programming and Language Programming languages derive much of their power from formally specified semantics. The language implementation, not the programmer, defines what if , while and return mean. I cannot persuade the compiler that false should be treated as true . The rules establish a shared and mechanically enforced understanding of what a program means. Large Language Models however, do not execute according to fixed semantics. They interpret natural language through context. This distinction has profound consequences and suggests that a language model has no intrinsic notion of authority. In a programming language, when two instructions conflict, the language specification and execution environment determine the outcome. In natural language, authority does not arise from the words alone. It depends on context, convention, identity, and external rules. Language models, by nature, inherit this ambiguity. A prompt is therefore not a program in the traditional sense. It is an attempt to establish the context within which subsequent language should be interpreted. "You are a detective." "Do not reveal the identity of the murderer." "Only answer questions using the evidence you have observed." None of these statements is mechanically enforced merely because it appears in the prompt. They describe a role, a constraint, and an assumed world. The model may follow them, but their authority must be created and protected by systems outside the model. Prompt injection exploits precisely this weakness. It

If you want to stream local media, this free and open source media server is just as good as Plex. But if you rely on remote access or live TV, prepare to tinker.

Demis Hassabis thinks the world needs an AI watchdog with the power to hit the brakes if frontier models become too dangerous. Writing in a blog post, the Google DeepMind CEO and cofounder said the US should lead the initiative, arguing that the country is the best place to set global standards "given its economic […]

Google released the Genkit Agents API in preview for TypeScript and Go. The open-source framework packages message history, tool loops, streaming, and state persistence behind a single chat() interface. Detached turns let agents work after clients disconnect. Interruptible tools provide human-in-the-loop control with anti-forgery validation on resume. By Steef-Jan Wiggers

Learn how enterprises can manage AI investments in the agentic era by measuring useful work per dollar, improving efficiency, and scaling high-value workflows.

When building multi-agent systems, rigid state graphs quickly fall apart in the face of dynamic user inputs. Imagine building a smart assistant: a user hands you a checklist of three household chores today, but tomorrow it might be a list of ten software debugging tasks. Because the number of tasks, their sequence, and their execution details are entirely runtime-dependent, you cannot hardcode this path at design time. Forcing dynamic lists of work into a static graph-based workflow can lead to fragile, over-engineered code. You need a workflow that adapts dynamically at runtime. The Google Agent Development Kit (ADK) provides a flexible programming model to define dynamic workflows . With the release of ADK 2.4.0 , triggering these workflows has become even more seamless: you can register a Workflow directly in an agent's tools list, allowing the coordinator agent to execute it automatically as a first-class tool. In this article, you learn how to configure and trigger a dynamic workflow directly from a coordinator agent. This guide uses a task list coordination example, but you can adjust this pattern to other dynamic orchestration needs. The architecture of a dynamic workflow Static workflows define the execution path at design time. Dynamic workflows, however, allow agents to invoke tools, spawn other nodes, and schedule sub-agents conditionally at runtime. The system consists of three main components: Root agent ( root_agent ) : Gathers the list of tasks from the user, requests final approval, and directly calls the tasks_workflow tool. The workflow ( tasks_workflow ) : A Workflow that iterates over the approved tasks. Sub-agent ( task_explainer ) : An Agent tasked with generating a step-by-step execution plan for each task. Here is the architectural diagram of the solution: Technical implementation Let's break down how to implement this solution using the Google ADK library in Python. The complete code resides in the devrel-demos repository with core logic in

A few weeks ago I was setting up a new project and needed to do the usual dance: create a Notion doc, spin up a Linear board, invite the team to Slack, and set up a couple of Zapier automations to connect them all. It took me most of an afternoon. That's when it hit me — I wasn't actually trying to "use" any of these tools. I just wanted the outcome. I wanted the project set up. And somewhere between the fifth Zapier trigger and the third failed webhook, I found myself thinking: why am I the one gluing all this together? That question is basically the whole thesis behind this post. AI agents aren't just a new feature category bolted onto SaaS. They're starting to eat the reason SaaS exists in the first place. The old deal: software rents you a workflow Traditional SaaS sells you a workflow, not an outcome. You pay for Notion, and Notion gives you a very nice, very rigid shape to pour your thoughts into. You pay for HubSpot, and it gives you a CRM shape. You pay for Zapier so you can awkwardly stitch the shapes together. This worked great for twenty years because the alternative was building everything yourself. SaaS was the shortcut. But the shortcut came with a tax: you had to adapt your work to fit the tool, and when you needed two tools to talk to each other, you had to become a part-time integrations engineer. The new deal: software does the workflow for you An AI agent flips that relationship. Instead of "here's a tool, go operate it," it's "here's the outcome, go figure out how to get there." You tell an agent "onboard this new client" and it can read the contract, create the folders, send the welcome email, schedule the kickoff call, and post a summary in Slack — using whatever tools it has access to, without you clicking through five different dashboards. That's the part that's easy to miss if you only think of agents as "chatbots with extra steps." A chatbot answers questions. An agent does multi-step work: It breaks a goal down into subtasks It calls tools

AI makes the first 80% of development feel fast, but hides architectural complexity until it's too late. To prevent system instability, engineering leaders must shift from raw throughput to systemic comprehension. By unifying spec-anchored SDD, TDD, and automated fitness functions into a repo-bound "Context Store," teams can ensure AI agents and human reviewers evolve code safely. By Stella Berhe, Stephan Bragner, Vikram Maran, Anand Jayaraman

New hyperscale data centers can't set up shop in New York for up to a year now that Governor Kathy Hochul (D) has signed the nation's first statewide moratorium. But a bill passed by the state legislature that could restrict even more developments still awaits her signature. The order blocks new environmental permits for data […]

Chrome's Gemini integration has arrived in the UK.

Drawing from the enduring adaptability of HTML and HTTP, Seph Gentle proposes embedding self-contained schemas directly into file headers, ensuring data remains readable without external definitions. His experimental format prioritises forward, backwards, and sideways compatibility, enabling data format evolution without central coordination or data loss By Olimpiu Pop

I Wish I Ran the Numbers on Open Source AI APIs Sooner Three months ago I would have told you self-hosting was the obvious move. "Open source means free, right?" I said that to a client while quoting them $3,500 for a GPU server setup. They smiled politely and went with someone else. That rejection sent me down a rabbit hole I wish I'd started years earlier, because the actual math — not the vibes-based math freelancers like me tend to do — completely flips the script. If you're running a solo practice or a tiny shop, you probably bill every minute of GPU babysitting straight out of your own pocket. That's time you could be shipping features, pitching clients, or — if we're being honest — sleeping. So let me walk you through what I learned the hard way, with all the pricing left exactly where it belongs. The Open Source Lineup That Actually Matters Right Now When I started this research, I assumed "open source AI API" was an oxymoron. If you're calling an API, somebody owns the server, so what's even the point of being open? Turns out the point is massive: open-weight models accessible through an API give you the pricing transparency of self-hosting without the DevOps funeral you're planning for your weekends. Here's the pricing matrix I put together from Global API's public rates. These are output token prices (input is usually cheaper), and yes — they're shockingly low compared to GPT-4o territory. Model License Output Price Self-Host Range DeepSeek V4 Flash Open weights $0.25/M $500-2,000/mo DeepSeek V3.2 Open weights $0.38/M $800-3,000/mo Qwen3-32B Apache 2.0 $0.28/M $400-1,500/mo Qwen3-8B Apache 2.0 $0.01/M $200-800/mo Qwen3.5-27B Apache 2.0 $0.19/M $300-1,200/mo ByteDance Seed-OSS-36B Open weights $0.20/M $500-2,000/mo GLM-4-32B Open weights $0.56/M $400-1,500/mo GLM-4-9B Open weights $0.01/M $200-800/mo Hunyuan-A13B Open weights $0.57/M $300-1,000/mo Ling-Flash-2.0 Open weights $0.50/M $300-1,000/mo Look at Qwen3-8B and GLM-4-9B at $0.01/M output tokens. A mi

Every business wants more leads. But the real challenge isn't generating them—it's identifying which leads deserve your team's attention first. Instead of manually reviewing every inquiry, we can build a simple AI-powered API that analyzes incoming leads and assigns a priority score automatically. In this article, I'll show a lightweight production-ready approach using Next.js 15 and Gemini 3.5 Flash. Project Structure app/ ├── api/ │ └── qualify/ │ └── route.ts ├── lib/ │ └── gemini.ts └── page.tsx API Route import { NextResponse } from "next/server"; export async function POST(req: Request) { const { company, message } = await req.json(); const prompt = ` Company: ${company} Message: ${message} Give: - Score (1-100) - Priority - Reason `; // Call Gemini API here return NextResponse.json({ success: true, score: 92, priority: "High" }); } Example Response { " score " : 92 , " priority " : " High " , " reason " : " Large company with a clear automation requirement. " } Now your CRM, chatbot, or automation workflow can instantly decide which leads should be contacted first. Why This Matters A simple AI scoring layer can help teams: Reduce manual lead review Respond faster to high-value prospects Prioritize enterprise customers Improve sales efficiency Save hours every week The best part is that this API can be connected to forms, chatbots, CRMs, or n8n workflows without changing your existing process. Production Tips Before deploying this to production, make sure you: Validate incoming requests Store API keys securely Add rate limiting Log AI responses for monitoring Cache repeated requests where appropriate Small improvements like these make a huge difference once traffic starts growing. Final Thoughts AI shouldn't replace your sales team—it should remove repetitive work so they can focus on conversations that actually matter. A lightweight lead qualification API is one of the fastest AI features you can add to an existing product, and it scales well as your business

Most automation advice assumes you're willing to pay for Zapier or spend weeks learning n8n. The business-automation-architect skill by @1kalin takes a different angle: your AI agent is already capable of running workflows on its own, using cron jobs, scripts, and built-in reasoning. No third-party automation platform required. The Core Premise Your agent has access to APIs, file systems, schedulers, messaging channels, and web tools. That's everything you need to automate business processes without installing anything else. The skill teaches you to think like an automation architect — finding the highest-value processes to automate, designing the workflow, implementing it with agent tools, and measuring the return. The philosophy is grounded: only automate processes that happen at least five times per week OR cost more than thirty minutes per occurrence. Below that threshold, the automation overhead rarely pays off. The 5x5 Automation Audit The first phase is a structured discovery process. The skill provides a scoring matrix across five dimensions — frequency, time cost, error impact, complexity, and number of systems involved. Each dimension is scored 0-3, giving a maximum score of 15. Processes scoring 12 or above are immediate candidates. Those between 8-11 go into the next sprint. Anything below 8 is left manual. The discovery questions are worth asking directly: what breaks when someone is sick? Where do things pile up waiting for a person? What data gets copied between systems every day? These are the real automation opportunities, and they rarely show up in generic automation advice. Designing the Workflow The skill defines a clear workflow architecture template covering triggers, inputs, steps, error handling, outputs, and monitoring. The trigger types supported are schedule (cron), webhook, event, manual, email, and file-based. Steps can be fetch, transform, send, decide, wait, or notify — each mapping directly to what an agent can actually do. Error hand

Changing the model for a running AI task is not a settings update. It is a distributed operation: read current task -> prepare credentials/config -> request restart -> receive result -> persist active model If two switches overlap, completion order can differ from request order. The system needs a rule for which intent wins. The concrete case At commit c58bcd4 , MonkeyCode records model-switch attempts with from/to model IDs, request ID, load-session flag, success, message, session ID, and timestamps in TaskModelSwitch . The reviewed task use case creates a switch record, asks taskflow to restart with the target model configuration, and completes the switch record and task model based on the response. The accompanying tests cover success and failure paths. From this source review, I could not establish an explicit compare-and-swap generation or a per-task serialization contract around overlapping requests. That does not prove an exploitable race: serialization may exist elsewhere in the deployment or taskflow boundary. It means concurrency semantics deserve an explicit test and contract. Why last completion is unstable Assume request A selects model A, then request B selects model B: time -> A: request ---- restart ---------------- complete B: request -- restart -- complete If each successful completion writes its model, B applies first and late A overwrites it. Reverse network timing and the result changes. The companion simulator makes that order dependence visible: export function naiveCompletionOrder ( completions ) { let model = " initial " ; for ( const completion of completions ) { if ( completion . success ) model = completion . model ; } return model ; } [A, B] ends on B. [B, A] ends on A. The caller's latest intent is not part of the rule. Add a monotonic generation Assign a generation while accepting each request: A -> generation 41 B -> generation 42 Completion may update active state only when its generation equals the task's current requested generatio

LLM Evaluation System Prompts Scored Rubrics Runtime Guardrails: A Practical Guide for Production Learn how to evaluate LLM outputs in production using system prompts, scored rubrics, and runtime guardrails to prevent hallucinations and ensure quality. TL;DR: To evaluate LLM outputs in production, combine system prompts that define evaluation criteria, scored rubrics using LLM-as-a-judge for dimensions like correctness and relevance, and runtime guardrails that filter or flag unsafe outputs. This approach scales better than human review, adapts via prompt changes, and catches failures that status codes miss, as seen in the Air Canada chatbot case. Why Production LLM Evaluation Demands More Than Status Codes A 200 status code only confirms the server processed the request—it says nothing about whether the generated text is factual, safe, or useful. The Air Canada chatbot that invented a non-existent bereavement discount returned perfectly valid HTTP responses, yet the hallucinated policy led to a tribunal ruling against the airline. Production evaluation must therefore separate operational health (latency, error rates) from output quality (correctness, relevance, harmlessness). Consider a typical API call that succeeds operationally but fails qualitatively: import requests response = requests . post ( " https://api.example.com/v1/chat " , json = { " model " : " gpt-4o " , " messages " : [{ " role " : " user " , " content " : " What is Air Canada ' s bereavement policy? " }]}, headers = { " Authorization " : " Bearer $KEY " } ) print ( response . status_code ) # 200 print ( response . json ()[ " choices " ][ 0 ][ " message " ][ " content " ]) # Output: "Air Canada offers full refunds for bereavement-related cancellations..." A 200 status code and a well-formed JSON body mask a completely fabricated policy. To catch this, you need a separate evaluation layer that scores the output against a rubric. LLM-as-a-judge is a common approach, using a second model to assess the

An activity log tells us what an agent did. A decision log should also tell us what it considered and rejected. Without rejected options, a later reviewer sees a clean path that never existed: model B was selected, the task restarted, the result succeeded. Missing are the reasons model A was unsuitable, why staying put was worse, and what new evidence would change the choice. That information matters for trust and recovery. It lets people challenge a decision without reconstructing the entire session. Execution history is necessary, but different The MonkeyCode model-switch record at commit c58bcd4 stores the task and user, from/to model IDs, request ID, whether to load the session, success, message, session ID, and timestamps. The switch use case creates that switch record, restarts the task with the target configuration, and records the result. That is valuable execution history. It answers “what switch was requested and what happened?” The expanded rejected-options structure below is my design proposal , not a claim about MonkeyCode's current schema or interface. Add the decision before the outcome A reusable record can separate choice from execution: { "decision_id" : "task-42-model-switch-7" , "context" : "The task needs the required tool-call contract." , "chosen" : { "option" : "model-b" , "reason" : "Passed the declared capability contract" , "evidence" : [ "evaluation/capability-model-b.json" ] }, "rejected" : [ { "option" : "model-a" , "reason" : "Required tool-call case failed" , "evidence" : [ "evaluation/capability-model-a.json" ], "revisit_when" : "Adapter version changes" } ], "execution" : { "request_id" : "req-switch-7" , "result" : "success" , "session_id" : "session-9" } } The key field is revisit_when . “Rejected” should not mean universally bad. It should mean unsuitable under a specific context and evidence set. Design the interface for progressive disclosure Do not paste this JSON into the main task timeline. Use three layers: Timeline: Switch

“On-device AI saves battery” and “cloud AI is more efficient” can both sound plausible. Neither is a measurement. The placement decision crosses at least four different budgets: user wait + network transfer + provider spend + device energy Do not collapse them into one vague “cost” number. Measure each with its own unit and evidence boundary. Start by identifying the actual execution path I reviewed MonkeyCode mobile code at commit c58bcd4 . The task stream opens a server-supported WebSocket. The speech-to-text hook also participates in a server-supported streaming path. That reviewed path is not evidence of on-device model inference. So a fair current study would measure a mobile client using remote task and voice services. An on-device alternative would be a separate prototype with its model, runtime, and packaging declared. Record a measurement envelope The included CSV template begins with these fields: sample_id,sample_kind,placement,device,os,framework,model,network,input_tokens,output_tokens,latency_ms,bytes_up,bytes_down,energy_joules,cost_usd Why so many? device , os , and framework make thermal and runtime results interpretable; model and token counts keep workload size visible; network separates offline, Wi-Fi, and cellular behavior; latency is milliseconds, transfer is bytes, energy is joules, and provider spend is currency; sample_kind prevents synthetic examples from masquerading as device measurements. Battery percentage is too coarse for short runs. It is affected by display, radio, background work, battery health, temperature, and OS estimation. If you cannot collect energy with an appropriate platform profiler or external power measurement, leave energy_joules empty. Use matched user flows Compare the same tasks, not unrelated model demos: Flow Cloud case On-device case Short prompt Same input and output cap Same semantic task and cap Voice turn Same audio fixture Same audio fixture Offline Expected failure or queued action Local completion if supp