01
WEB PLATFORMS & SAAS
Accounts, roles, payments, dashboards. Products built to survive real users, not just launch day.
BLAST MONSTER — development studio
> DEVELOPMENT STUDIO
WE EAT HARD PROBLEMS
BLAST MONSTER is a development studio for the builds other teams quote nervously: high-load backends, AI agents, Telegram bots and Mini Apps, mobile apps and 3D web.
FEED THE MONSTER SEE WHAT WE ATE
Relax. It only bites on click.
CLICK ANYWHERE — IT BITES ·
01 / TEETH
What we chew with. Sharpened daily.
TypeScript, React, Next.js, Vite, three.js, WebGL, GLSL, GSAP, React Native, Swift, Kotlin, Figma, Telegram Mini Apps, Python, Go, Node.js, FastAPI, aiogram, Telethon, PostgreSQL, ClickHouse, Redis, Celery, Qdrant, LLM APIs, RAG, n8n, Docker, Sentry
02 / THROAT
Most studios bite off only what they can chew.
We hunt the rest: load that melts servers, a thousand bots in one process, AI that has to work on Monday, not just in the demo.
We chew it all the way to production.
And we’re still hungry.
— BLAST MONSTER, 2026
03 / DIET
WHAT WE EAT
Seven dishes on the menu. The harder, the tastier.
02
AI AGENTS & AUTOMATION
LLM agents that read documents, answer clients and fill in the CRM. Actual work, not a chatbot for the pitch deck.
03
TELEGRAM BOTS & MINI APPS
From a single bot to a fleet. Our Telegram HR-CRM handles 1000+ bots in one process, so yours won’t scare us.
04
HIGH-LOAD BACKEND
APIs, queues and data pipelines that stay boring under pressure. Boring is the whole point.
05
MOBILE APPS
iOS and Android apps with a real backend behind them: push, payments, offline, the lot.
06
3D & WEBGL SITES
Sites like this one: scroll-driven 3D that loads fast and doesn’t cook your phone.
07
PRODUCT DESIGN
UX, UI and design systems from people who also build them. Nothing gets lost between Figma and production.
Not on the menu? Bring it anyway. We’ll have a taste. FEED THE MONSTER
04 / DIGESTED
DIGESTED
Proof of appetite. Real systems, not mood boards: what the problem was, what it runs on, what it does now.
-
#01 · TELEGRAM
TGBLAST
SaaS · Telegram conveyor
IN PRODUCTIONYEAR 2026
> PREY
Telegram outreach usually lives in a zoo of scripts, spreadsheets and burned accounts. Teams needed one cloud cabinet that runs 24/7.
> WHAT WE DID
- Account warm-up with anti-spam pauses, checker, proxy pool with cooldowns
- Parsing in four modes, batch campaigns with daily limits, unified inbox
- Stories and reactions, link tracking, AI commentator on the client’s own key
> RESULT
A live product with public pricing and a free trial. One cabinet, all tools, zero zoo.
-
#02 · TELEGRAM
HR HELPER
HR-CRM inside Telegram
IN PRODUCTIONEATEN IN PRIVATE · NDAYEAR 2026
> PREY
Hiring through Telegram at scale: every vacancy is its own bot, every candidate needs a questionnaire and a live answer from HR.
> WHAT WE DID
- One webhook for all bots, FSM questionnaires, a forum topic per candidate
- Two-way relay: HR answers from the topic or the web admin over a live SSE chat
- Roles, TOTP 2FA, encrypted bot tokens, funnel analytics in ClickHouse
> RESULT
Built for 1000+ bots in one process. Message order holds even under Telegram rate limits.
-
#03 · HIGH-LOAD
FLEET CRM
CRM for a fleet of Telegram accounts
IN PRODUCTIONEATEN IN PRIVATE · NDA
> PREY
Thousands of Telegram accounts, proxies and campaigns run by several operators — without two processes ever grabbing the same account.
> WHAT WE DID
- Bulk import (.session / tdata / Google Drive) with automatic health checks
- Celery queues for warm-up, campaigns, group checks and LLM comments
- Distributed locks and auto-recovery of stuck jobs; per-operator pools and roles
> RESULT
6 automation modules in one panel, multi-operator by design.
-
#04 · AI AGENTS
CONSUL
Legal AI for cross-border money
IN PRODUCTIONEATEN IN PRIVATE · NDA
> PREY
Advice on taxes, currency control and moving funds across 5 jurisdictions — with references to real norms and no sensitive data leaking to the cloud.
> WHAT WE DID
- A local Qwen 32B anonymises the request and reads documents inside a closed perimeter
- Only clean text goes to Claude for a grounded answer from the norms base
- RAG over Qdrant filtered by jurisdiction; the base refreshes itself every 24 h
> RESULT
237 legal acts, ~37,000 chunks of norms, 5 jurisdictions. Fresh every 24 hours.
-
#05 · AI AGENTS
COMMENTATOR 3000
Channel activity automation
IN PRODUCTION
> PREY
A channel post needs a discussion the moment it goes live — with replies that actually match the post.
> WHAT WE DID
- A channel monitor triggers a scenario on every new post
- An LLM writes comments and replies matched to the post topic
- Account pool with proxies, managed from a web panel over HTTPS
> RESULT
A pool of 20,000 accounts reacts as soon as a post is published.
-
#06 · AI AGENTS
TGSEARCH
Multi-tenant Telegram marketing platform
IN PRODUCTION
> PREY
Find the right Telegram chats for a topic, analyse them and run campaigns — for users who must never see each other’s accounts.
> WHAT WE DID
- 1000 Celery queues: chat search, GPT analysis with tgstat enrichment, campaigns
- Fully isolated pools per user: accounts, proxies, ad lists
- FloodWait handling and account health monitoring
> RESULT
1000+ users, 1000+ isolated pools, 1000 queues, one shared admin.
05 / THE HUNT
THE HUNT
Four moves. Same every time. That’s why they work.
-
01
TRACK
We sniff out the real problem behind the brief and ask the uncomfortable questions now, not in month three. You get a fixed scope, a price and a plan.
-
02
BITE
First sprint, first working slice. The architecture is set, the repo is open, and the demo is something you can click, not a promise you can frame.
-
03
CHEW
Sprint after sprint we grind through the backlog. There’s a demo at every stage, so you always see what’s been eaten and what’s next.
-
04
BLAST
Launch to production with monitoring, backups and alerts wired in. Then we stick around, because the monster guards what it builds.
06 / CORE
FEED THE MONSTER
Drop your problem into the core. The harder it is, the brighter it glows. A human replies. Ironic for a studio that builds bots, we know.