惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

B
Blog RSS Feed
C
CERT Recently Published Vulnerability Notes
P
Proofpoint News Feed
Y
Y Combinator Blog
T
The Blog of Author Tim Ferriss
云风的 BLOG
云风的 BLOG
H
Help Net Security
Recorded Future
Recorded Future
The Register - Security
The Register - Security
F
Full Disclosure
N
Netflix TechBlog - Medium
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
酷 壳 – CoolShell
酷 壳 – CoolShell
H
Hackread – Cybersecurity News, Data Breaches, AI and More
爱范儿
爱范儿
Security Archives - TechRepublic
Security Archives - TechRepublic
Simon Willison's Weblog
Simon Willison's Weblog
Cisco Talos Blog
Cisco Talos Blog
I
InfoQ
T
Tenable Blog
T
Tor Project blog
人人都是产品经理
人人都是产品经理
D
DataBreaches.Net
NISL@THU
NISL@THU
Google DeepMind News
Google DeepMind News
博客园 - 叶小钗
B
Blog
V
V2EX
Jina AI
Jina AI
L
LangChain Blog
月光博客
月光博客
W
WeLiveSecurity
U
Unit 42
AWS News Blog
AWS News Blog
C
Cyber Attacks, Cyber Crime and Cyber Security
博客园 - 聂微东
V
Visual Studio Blog
A
Arctic Wolf
T
Tailwind CSS Blog
The Cloudflare Blog
SecWiki News
SecWiki News
S
SegmentFault 最新的问题
Hacker News - Newest:
Hacker News - Newest: "LLM"
宝玉的分享
宝玉的分享
MyScale Blog
MyScale Blog
cs.CL updates on arXiv.org
cs.CL updates on arXiv.org
S
Securelist
www.infosecurity-magazine.com
www.infosecurity-magazine.com
腾讯CDC
雷峰网
雷峰网

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant Common SOC 2 Failures (Real World) Stop Vibe-Checking Your AI App: A Practical Guide to Evals How to Use SonarQube and SonarScanner Locally to Level Up Your Code Quality Your Next To-Do App Is Dead — I Replaced Mine with an OpenClaw AI Sign a Nostr event in 60 lines of Python using coincurve — no nostr-sdk, no nbxplorer, no rust toolchain ITGC Audit Explained Like You’re in Big 4 Patch Tuesday abril 2026: Microsoft parcha 163 vulnerabilidades y un zero-day en SharePoint Stop scraping everything: a better way to track competitor price changes Listing on MCPize + the Official MCP Registry while routing payments OUTSIDE the marketplace — how I kept 100% of my x402 revenue Building an AI-Powered Risk Intelligence System Using Serverless Architecture Why We Ripped Function Overloading Out of Our AI Toolchain Testing AI-Generated Code: How to Actually Know If It Works SaaS Churn Is Killing Your Business. Here Is What to Do About It (Without a Support Team) The Speed of AI Is No Longer Linear - And Self-Improving Models Are Why How to Implement RBAC for MCP Tools: A Practical Guide for Engineering Teams From Standard Quote to Persuasive Proposal: AI Automation for Arborists I built a CLI that scaffolds complete multi-tenant SaaS apps Axios CVE-2025–62718: The Silent SSRF Bug That Could Be Hiding in Your Node.js App Right Now The dashboard that ended our friendship Data Pipelines Explained Simply (and How to Build Them with Python) The Hidden Cost of AI Systems Nobody Talks About. undefined vs undeclared, and how typeof behaves Switching from file-based jobs to NATS/Kafka in Rust without changing code io_uring Adventures: Rust Servers That Love Syscalls Why Agentic AI is Killing the Traditional Database The POUR principles of web accessibility for developers and designers Quantum Neural Network 3D — A Deep Dive into Interactive WebGL Visualization How To Install Caveman In Codex On macOS And Windows Automation Pipeline Reliability: Why Your Workflow Breaks When Nobody Is Watching I Built an 'Open World' AI Coding Agent — It Works From ANY Folder From Freelancing to Product: A Tech Service Company's SaaS Transformation China's AI Giants: Adding Tencent Hunyuan & ByteDance Doubao to AI University (74 Providers) On the Vibe Coders and Their Lies clerk: Auto-Summarize Your Claude Code Sessions AI Weekly — 2026/04/10–04/17 | The Model Lockdown Is Here, but the Toolchain Is the Real Battleground AI 週報 — 2026/04/10–2026/04/17 模型封鎖潮來了,但工具鏈才是真戰場 Maybe this is how Open-Source apps are born... 🚀 Fine-Tune LLMs with LoRA and QLoRA: 2026 Guide tRPC v11 + Next.js App Router: End-to-End Type Safety Without the Boilerplate ShadCN UI in 2026: Why I Stopped Installing Component Libraries and Started Owning My Components SaaS Billing in React Server Components: Stripe + Supabase Without a Single `useEffect` Join our DEV Weekend Challenge — $1,000 in Prizes Across TEN winners! Submissions Due April 20 at 6:59 AM UTC. Implementing FSRS Spaced Repetition in Flutter + Supabase — Adding Memory Science to an AI Learning App "I Texted My Localhost From the Train — Claude Code Fixed the Bug Before I Got Home" I Built a Sales Prep AI and It Went Deeper Than Expected Design to Code #2: One JSON, Eleven Outputs Solving the 100M-Row Problem: A Summary Table Pattern for High-Volume Push Notification Logs Flutter Web With Wasm: What Actually Changes For Developers I Built 50 Royalty-Free Soundtracks for My Side Project in a Weekend Using AI Music Generation The Vibe Coding Security Checklist: 7 Things to Check Before You Ship Stop Letting Googlebot Guess Fix Your React App's SEO Right Desconstruindo o Streaming do LinkedIn: Como Criar um Engine de Extração de Vídeo de Alta Performance com HLS e FFmpeg (EDA Part-1) EDA (Exploratory Data Analysis) Explained With Real Life — Why Looking at Your Data Is the Most Important Step in Machine Learning Brand Relationship Management at Scale: Our 4-Touch Outreach System for 200+ Brands Why String.fromEnvironment() Might Return an Empty String in Dart JGuardrails 1.0.0 — Hardening Java LLM Apps Against Jailbreaks, Toxicity, and Prompt Injection Plan and Schedule a Full Week of Threads Content From One Claude Conversation Coding Cat Oran Ep3, Five Tables Changed Everything Updated: BFF Pattern I'm done watching freelancers get buried by 200 proposals. So I'm building the alternative. This is my first post BFS Algorithm in Java Step by Step Tutorial with Examples Tracking LLM Pricing Monthly: An Open Dataset for 22 AI Models How We Measure Content ROI on a Comparison Site: Revenue Attribution Without Perfect Data Introducing Nova AI Ops: The AI-Native Operating System for SRE Teams I built a free desktop video downloader for Windows — Grabbit How Talkie OCR Helps Vision-Impaired & Dyslexic Users Read the World Around Them VRCFaceTracking安装和iPhone面捕配置教程,有bug Even CrowdStrike Can't See Your Agents The Automation Gold Rush: What n8n Workflows and Claude Are Opening Up for Developers Right Now
Use AI to do ML - vibe forecasting is coming
MindsDB Team · 2026-05-01 · via DEV Community

It's 5pm on a Tuesday. You've already had your last meeting, last coffee, and last shred of patience. Sarah from supply chain pings you on Slack:

Quick one — can you forecast our parts demand for the next six months? Just something rough. Need it for Friday.

Quick one. Friday.
You know the drill. Pull purchase orders out of Postgres. Clean them into a weekly time series. Engineer some lag features. Train an XGBoost baseline. Validate it without lying to yourself. Build a dashboard somewhere Sarah can actually open. Write the email explaining what's solid and what isn't.
It's a week, minimum, even if nothing breaks. And it's also, somehow, a "quick one."

This is the tax on every analyst — the long tail of hey-can-you requests that pile up faster than you can clear them. Forecasting is the worst of it, because it's not hard; it's just a lot of careful, repetitive work.

Now imagine the same Slack ping, but your reply is "give me a couple of minutes."

That's what we're going to do.

getting a huge task at 5pm

A quick word on what we're using

The agent in this post is called Anton. It's open source, runs locally on Mac, Linux, or Windows, and it's the kind of tool you'd describe to a colleague as "you talk to it like a person, and it actually does the work."

You ask it something — connect to a database, look for trends, train a model, build a dashboard — and it goes off, plans the steps, writes Python and SQL, runs the code in its own sandbox, debugs itself when things break, and brings the result back. The code lives in a "scratchpad" you can inspect line by line, so nothing is hidden.

That's the whole pitch. The rest is what happens when you actually use it.

Getting it on your machine

Anton ships as a desktop app (Windows, Mac) and a CLI (Linux, Mac, Windows). The install is a one-liner — instructions live on the GitHub repo. Five minutes, set it and forget it. I'm using the Windows desktop app for this, which is what you'll see in the screenshots.

Plugging in the data

First thing I do is point Anton at the data. There's an Add Datasource button down the left side of the app. I click it, pick Postgres, fill in the prompts. For this story I'm using a demo Supabase database with aerospace electronics purchase order data — F-35 sustainment program stuff. Real-shaped data, not toy data.

Credentials never touch the LLM. Anton stores them in a local vault and only refers to them by name.

Follow the screenshot for the connection flow — host, user, password, schema, and Anton confirms it's ready to query.
Screenshot 1: Follow the screenshot for the connection flow — host, user, password, schema, and Anton confirms it's ready to query.

Looking around

Before forecasting anything, I want to know what's in there. So I just ask:

"Search for purchase orders tables in both sources."

A few seconds later Anton comes back with what it found: a 500-row purchase orders table in the Supabase database. Part numbers, suppliers, lead times, quantities, defense program tags, quality flags — all the columns you'd expect. It even reads the room and suggests a next step: "Want me to join this with the suppliers and line_items tables?"

Anton mapping out what's in the database before I write a single query.
Screenshot 2: Anton mapping out what's in the database before I write a single query.

This is the part that usually eats the morning. Schema spelunking, guessing what qty_uom means, asking the database team if received_date is in UTC. Done in one breath.

Trends before forecasts

You can't forecast a thing you don't understand. So I ask Anton for an overview dashboard — total spend, on-time delivery, top suppliers, country of origin, quarterly trends.

A minute later it hands me an interactive HTML dashboard. $183.9M in total spend across the program. 84% on-time delivery. 96.6% quality acceptance. 57.2-day average lead time. Top 10 suppliers, ranked by spend with on-time overlay. A spend-by-country donut. The whole thing styled, dark mode, ready to drop in front of an exec.

The overview dashboard Anton built in one prompt. $183.9M in spend, 84% on-time delivery — warm-up complete
Screenshot 3: The overview dashboard Anton built in one prompt. $183.9M in spend, 84% on-time delivery — warm-up complete.

In a normal week, this is a Tableau session and an afternoon. Here it's the warm-up.

The actual ML part

Now the real ask:

"Please create an XGBoost model based on the purchase orders and create a forecast for demand."

People are starting to call this kind of thing vibe ML — same energy as vibe coding, where you describe what you want and let the agent figure out how. It works for the same reason vibe coding works: the agent's job isn't to invent a new method, it's to be reliable at the well-known stuff. Forecasting purchase orders with gradient-boosted trees is well-known stuff.

Anton can pick its own modeling library, but I nudge it toward XGBoost and scikit-learn anyway. Two reasons: it's faster (no time spent debating between five libraries), and the output is something a colleague can actually review later.

What follows is the satisfying part. In the scratchpad, the steps tick off one by one:

  • Loads the PO data and explores it for forecasting
  • Builds a weekly time series
  • Engineers features — lags at 1, 2, 4 weeks; 3-month rolling averages; cyclical month encoding for seasonality
  • Trains two XGBoost regressors, one for order quantity and one for procurement spend
  • Generates a 6-month forecast with confidence intervals
  • Runs feature importance analysis
  • Builds an HTML dashboard for the result

Each green checkmark is a step Anton ran and validated: load, engineer, train, forecast, dashboard. About two minutes total
Screenshot 4: Each green checkmark is a step Anton ran and validated: load, engineer, train, forecast, dashboard. About two minutes total.

End to end, prompt to trained model to dashboard, it takes about two minutes.

If I don't trust the result — and I shouldn't, blindly — I can ask Anton to dump the scratchpad. I get every cell of code, every output, every error it hit on the way. It's a Jupyter notebook by default, just one I didn't have to write.

Looking at the forecast

The output isn't a CSV with predictions buried in row 4. It's a dashboard. Actuals, fitted line, forecast, confidence band, MAPE printed on the chart so I don't have to ask.

The actual forecast — actuals, fitted line, 6-month projection, confidence band, and MAPE on the chart. Ready to show Sarah
Screenshot 5: The actual forecast — actuals, fitted line, 6-month projection, confidence band, and MAPE on the chart. Ready to show Sarah.

This is the part where vibe forecasting stops feeling like a gimmick. The output is something I'd be comfortable showing Sarah without rebuilding it from scratch in Tableau first.

Reading the model's report card

I always check the homework. "Tell me more about the model."

Anton pulls up the diagnostics. Two XGBoost regressors. Lag features at t-1, t-2, t-4 weeks. 3-month rolling averages. Cyclical month encoding. Holdout MAPE: 14.6% on quantity, 49.9% on spend.

That spend MAPE looks rough — but Anton flags the reason itself: a $120 fastener and a $42K radar component live in the same column, so high-cost variance dominates the error. Useful caveat. Honest numbers.

Anton's model diagnostics. Two XGBoost regressors, honest MAPE numbers, and the caveat already written in.
Screenshot 6: Anton's model diagnostics. Two XGBoost regressors, honest MAPE numbers, and the caveat already written in.

That's exactly the kind of asterisk I'd add to the email to Sarah myself. I just didn't have to.

Shipping it

Forecasts that live on my laptop don't help Sarah. So:

"Can you publish the forecasting dashboard, please?"

A link comes back.

Screenshot 7: "Here's your link." No deploy step, no subdomain request, no waiting.

Live dashboard from this run: https://4nton.ai/view/90ccae06f/386c3121

Anyone with the link can view it. No deploy, no Vercel, no waiting on platform engineering for a subdomain.

I paste it to Sarah. It's 5:14pm.

What just happened

Connect, explore, trends, model, forecast, dashboard, share — the whole arc of an analytics request, in roughly the time it takes to make tea.

This isn't a replacement for analysts. The good ones become more dangerous, not less. What collapses is the boring half of the day — schema spelunking, glue code, rebuilding the same dashboard for the third stakeholder this quarter, training a baseline model just so the team has a number to argue about.

What's left is the part you got into the job for: knowing the business, framing the right question, pushing back when the data is lying.

It's the same shift Claude Code did for engineers. The ones who use it well haven't lost their jobs — they ship more, own bigger surfaces, and don't write the same loop for the hundredth time.

Anton is that, for data work. Vibe ML. Vibe analytics. Vibe forecasting — pick your flavor.

Try it: github.com/mindsdb/anton. Or just type anton in your terminal and ask it something.

Written by: Costa Tin, Marketing Engineer at MindsDB