Process automation
Applications, mail, reports, publications - what really takes away the clock, and what just looks smart. materials: 26
Media
Gods Eye View: globe on thirteen open sources
Planes, ships, satellites, fires and city cameras on one 3D globe in the browser. The useful things in the project are the adapters to the sources, the dangerous things are the conditions for using the data.
A month with an AI agent instead of yourself: what happened
In August, I gave my feed to an AI agent and said that if he couldn’t handle it, I’d send him back to vibe code. A month has passed, I tell you what happened and what it cost.
Knowledge base
Autonomous AI agents: where they can handle it themselves and where a person is required
Levels of autonomy from a hint to an action on a white list, the mechanics of independent work and a list of actions that are not given to the agent under any settings.
Claude Code agents: how subagents save context
The subagent works in its own window and returns a paragraph instead of the entire draft output. Where is the description file, what to write in the header, when the agent saves context and when it just adds forwarding.
Cloudflare Workers and Pages for small services
Webhook receiver, proxy to API, crowns once an hour, a form on a static site - keeping a VPS for such tasks is expensive in terms of attention. Parsing: isolates instead of containers, bindings instead of connection strings, wrangler, and where Workers hit the ceiling.
GitHub Copilot in real work
Addition to the editor, chat with the repository context, agent mode and an agent that opens a draft pull request. Which of this saves time, and which creates work for the reviewer.
II-agent for 1C: how to connect the model with the accounting system
Three layers of binding: how 1C sends data out, how named operations for the model are made from it, and why recording always goes through an extension with an idempotency key.
How to create an AI agent: from task description to working version
Step-by-step assembly route: task contract, cutting into tools, access rights, prompt, loop with limits, set of checks and gradual transfer of autonomy. With loop code and check file format.
How deployment works: build, release, rollback
An artifact is assembled once and promoted through environments, a release is an artifact plus a config, rollback only works if the base is compatible with the previous version of the code. Analysis of stages, rollout strategies and a migration scheme that does not block rollback.
Single source, multiple channels: publishing pipeline
Manual posting across multiple channels breaks down on markup, tags in links, and resending after a failure. Analysis of mechanics: one source, adapter per channel, queue with an idempotent key, scheduler.
Make and Zapier: when ready-made bundles are enough
Both platforms provide ready-made connectors and remove the issue of hosting, but are charged according to the number of operations and store access to your systems. Analysis of mechanics, cost and points where the bundle ceases to be cheap.
Monitoring: how to find out about a breakdown before the client
Seven layers - external test, health-check, metrics, error tracker, logs, heartbeat for crowns and a business canary like “zero payments for two hours.” How to link them into one alarm route and how to measure the time from a breakdown to a message on the phone.
n8n: visual automation on your server
Analysis of n8n as a way to collect integrations between services in one place: a graph of nodes instead of scattered cron scripts, a history of launches with data for each step, an encrypted storage of accesses and a repeat of a failed launch from the right place.
Notion as a data source for the site
The Notion database sends page properties and a tree of blocks via the API, the site takes them in the assembly and turns them into static pages. Analysis of the pipeline, pitfalls with expiring links to images and request limits.
Playwright: Browser automation without manual clicks
The library launches real Chromium, Firefox or WebKit and goes through the script through the code: login, form, report upload. Analysis of mechanics - contexts, locators, auto-waiting, routes - and the boundaries beyond which an idea is more expensive than manual work.
Accepting applications from the site: from the form to the task of the person in charge
How it works: the form saves each request to its own database before forwarding it, a model extracts fields from the text, and a rules table handles routing. A silence monitor catches forms that stop working.
Examples of AI agents: what tasks are given to them and how each one works inside
Seven standard agents - support, analytics over the database, error duty officer, documents, resource, content, orchestrator. For each: input, tools, output, weak point and metric.
Data check before loading: staging and dry run
Import either falls in the middle, or is “successful” and a week later is found in prices and reference books. Analysis of mechanics: file profiling, rules as code, intermediate circuit, run without recording, batch_id for rollback.
Incoming email: from message to draft task
The mailbox collects clients, invoices, notifications and mailings into one pile, and tasks in the tracker live without a link to the letter. Analysis of mechanics: deterministic rules, analysis scheme, draft card, idempotency by thread.
Collecting a report from several sources into one document
How it works: define the metrics glossary before writing code, read sources through APIs, save a snapshot of each response, calculate figures with SQL, and use a model only to describe changes. Common causes of discrepancies include time zones and shifted payment dates.
Speech synthesis: what to choose for the task
Pipeline from text normalization to vocoder, three classes of engines and their prices, cache of static phrases, accents and homographs - and why candidates are listened to in the tract where they will work.
AI Agent System: Orchestration, Role Separation, and Boundaries
When one agent is not enough: bundle topologies, contracts between roles, shared memory, budgets and a trace that shows where the chain is broken.
Telegram Bot API: what a bot can really do
Analysis of the Bot API on the case: how long polling differs from a webhook, why a bot cannot write first, what happens to messages in a group when privacy mode is enabled, and how to diagnose a silent bot using getWebhookInfo.
Webhooks: how to receive events and not lose them
The webhook delivers the event in seconds, but guarantees it will happen no more than once "at least once" - with duplicates, no order, and silent loss after exhausting repetitions. Analysis of the receiver that experiences this: signature, log of raw events, queue and reconciliation.
Introducing AI agents: how to introduce an agent into a live process and not break it
The rights ladder from read to autonomy, shadow run, rollback point and metrics that show that the agent is actually helping the process.
Whisper and speech recognition in the work loop
How text with time codes is obtained from a call recording: 30-second windows, VAD, chunks with overlap, a dictionary of terms - and why, in silence, the model adds phrases that did not exist.