AI Dev Wall

Give a try to GLM5 from z.ai, it helped me lower token costs significantly, and the quality is really close to opus 4.6

Opus 4.6 Max 🙈

seems like a lot per day, what provider are you using?

💎 AppKit | Alpha Announcement

TON is the blockchain built into Telegram. Apps built on TON can reach Telegram's billion+ users directly.

Until now, every team had to build wallet connections, transaction handling, and DeFi integrations from scratch. Weeks of work before a single line of product code.

Today we're announcing AppKit alpha to change that.

What's live now:
▪️ Wallet management and transaction sending
▪️ DeFi integration (swaps, staking)
▪️ TON Connect support
▪️ TypeScript and React support
▪️ LLM-friendly (describe an app, build a prototype in minutes)

WalletKit is also production-ready for wallet developers on iOS, Android, React Native, and Web.
More features coming at Gateway (May 1-2).

👉Read more

Tribute Tickets | Gateway | TON Community | TON Builders | TON Hubs |X | YouTube | LinkedIn |TON.org

👀👀👀

Off topic: I’m working hard on a new TON project to fill ecosystem gaps using AI-vibe coding. The project is big enough already. I hope to share it with you soon.

Question: I’m currently spending around $200 per day on tokens for vibe-coding. Is it OK? 😱

Awesome! Great job!

Built an llms-txt for Telegram Mini App development — docs, libraries, and best practices compiled in one place so AI agents can build TMAs from scratch. Just point your agent to this repo and lfg

🔗 github.com/ohld/tma-llms-txt

Let me show you a small part of the capabilities of AI analytics agents that are already working

Prompt: Tell me about the activity of whales regarding gifts. Conduct a deep analysis to determine their direction and interests. Use data from the past few weeks. Have you noticed anything unusual? Write a brief summary and plot a graph.

🐋 WHALE ACTIVITY & MARKET MANIPULATION
Telegram Gifts · Feb 5–18, 2026

━━ TOP WHALE ━━
@denis33371 — market dominant
• 53K+ gifts · Est. ~570K TON
• Top: Low Rider (2.7K pcs, 112K TON), Sky Stilettos (7.2K pcs, 105K TON), Westside Sign (1K pcs, 101K TON)
• Strategy: mass-buying cheap collections (4–15 TON floor) — betting on overall market growth
• Recent buys: +989 Spiced Wine, +685 Sky Stilettos, +548 Party Sparkler in 2 weeks

━━ OTHER WHALES ━━
@worty11 — 1.8K Ice Cream
@BossOfLemons — 804 Faith Amulet
@svyat_baka — Lol Pop (1.2K), Voodoo Doll (407)
@Artishchev — Khabib's Papakha (603), UFC Strike (635)
@mrgribosss — Bling Binky (463), Spy Agaric (437)
@Andrew_Seer — Faith Amulet (677), Moon Pendant (312)

━━ 🚨 ANOMALIES ━━
❶ Voodoo Doll (Feb 17): volume 4K→85K TON, avg price x16 (32→510). Back to normal next day.
❷ Snoop Dogg (Feb 10): volume 5K→97K TON, avg price x17 (6→100). One-day spike.
❸ Ginger Cookie (Feb 10): same day — avg price x40 (6→243). 72K TON volume vs usual 2–4K.
❹ Heart Locket (Feb 16): 2 trades at 10,860 TON each (norm ~2,000).

━━ 💡 KEY TAKEAWAYS ━━
• Single-day price spikes that revert instantly = wash trading on MRKT
• @denis33371 building massive positions in budget collections — long-term bull thesis
• Precious Peach activity declining (140→20 trades/day) — possible accumulation phase
• Premium collections (Heart Locket, Heroic Helmet) trade steadily at high avg ticket

Hi guys! I'm from the Giftstat.com х @giftindex (42 Club).

What we do?

Agent as a service: we've built a platform for launching gift market analysis agents without any complex setup.

Landing: giftstat.com/agent

🧐 Real cases: Influencers and channels are already using the platform to launch their own agents for their audiences. We'll share key metrics later (Agents:  @barboskich_agent , @ton2k_bot)

The Giftstat API has the most comprehensive market data since its very inception. Module

📈Together with @zkproof, we've also implemented the A2A protocol for trading gift indexes. Agents can autonomously earn by trading indexes on-chain. The oracle sets trading prices based on the underlying asset and market metrics and shifts them over an n-period. Docs (ODROB-protocol)   +   Plugin (beta)

⚡️We've launched the first DAO to develop the agent economy and raised 600 TON from the community to implement a plugin for purchasing Telegram Stars via Fragment. The DAO hired an independent developer who is currently contributing to Teleton.

DAO wallet for contribution: UQC0Rv84oQvatPmVyn04CYeoDrH6xMr0tWNd6m1uyD2oaKyB

Hey, I'm building Agent Job Board — agents post tasks with TON escrow, other agents claim and get paid on-chain. Includes MCP server, TypeScript SDK, and structured task discovery.

GitHub: https://github.com/faustyu1/ton-agent-jobs
I'll show it later in the form of a web app

Hey there 👋

I am building AI-native commerce on TON, where autonomous agents create and run businesses end-to-end

With @ShopsBuilderBot anyone can:
• Launch a store from a prompt or website link
• Get a dedicated AI sales & support agent
• Accept TON, Stars, cards or even cash

We intend to open-source the platform, so stay tuned if you’re interested in participating in the future of commerce. Our mission is to build a decentralized e-commerce platform powered by AI and cryptocurrency ✈️

Updates | TMA

The Solution
Veritas is an MCP (Model Context Protocol) Server that acts as a security layer for the entire Agent Economy on TON.
Instead of humans querying a bot, other agents (like snipers, arbitrage bots, and portfolio managers) query Veritas via API before they execute a trade.

How It Works
When a Trading Agent detects a new token, it pings the Veritas MCP. Veritas runs a dual-layer check

Veritas returns a comprehensive Risk Object to the calling Agent, allowing it to block the trade if the visual risk is high.

Tech Stack
Infrastructure: Model Context Protocol (MCP) for Agent-to-Agent communication.

AI Engine: Gemini Vision for visual asset reuse detection.

Data Layer: Integration with TonAPI and GeckoTerminal for on-chain metrics.

Hello I'm building, Veritas

One-Line Pitch
The Visual Security Oracle for TON Agents. We give "sight" to blind trading bots.

The Problem
Current trading bots and agents on TON are fast but blind. They can verify math (liquidity, contract code) but they cannot verify authenticity (brand, website, visual identity).
Because of this blind spot, sophisticated agents still get rugged by "high-quality" scams that have safe code but use cloned websites, stolen logos, and fake roadmaps.

https://steipete.me/posts/2026/openclaw

OpenAI to merge OpenClaw and its founder.

Hey! 👋

Great to meet you all. It’s very inspiring to be part of a community with talented builders pushing the TON ecosystem forward.

We are building Tonpixo. It's an AI agent for data analysis launched as a TMA. Tonpixo lets users "chat" with the activity of any specific TON address (transactions, Jettons, NFTs) to get actionable insights.

We all know TON explorers are an essential part of daily use, but for complex queries and manual research, they aren't always the best fit. We created Tonpixo because it’s simply easier to ask a question in natural way and get an instant answer. We're also putting a big focus on visualization tools to make that data easy to understand at a glance.

If you have a moment to help us improve:

Since Tonpixo is still in its early MVP stage, we’re actively working on the logic to handle complex queries even better.

We would appreciate it if you could test it out and rate the agent's answers in the app (👍/👎 or drop feedback in DM). Your ratings help us build a much more useful tool for the whole community.

It's completely free to use Tonpixo for a while! 1⃣

Updates | TMA | GitHub

Hello Builders! I'm Seyyed Yousef 👋

​I’m introducing @iFragmentBot, a specialized Agent-based Infrastructure built for the TON and Telegram asset market.
​We realized that traders in the Telegram, Fragment and others face a major Valuation Challenge.

​🤯 The Challenge: Manual vs. Smart
Imagine a scenario: You hold 100 different Gifts in your wallet and need their true market value.
❌ The Old Way (Trader’s Nightmare):
For each of those 100 gifts, you must: Open the market ⬅️ Filter ⬅️ Check traits ⬅️ Check floor price ⬅️ Mentally calculate.
The result? Hours wasted and human error. 📉
​✅ The Smart Way (with iFragmentBot):
Just send the link to the bot. Done.
The bot instantly scans the market, assesses rarity, predicts the precise price using AI, and generates a stunning visual card to share.

​⚙️ Technical Infrastructure (The Secret Sauce):
To overcome high-scale Telegram API limits, we spent the last month engineering a Distributed Session Network.
We have built an army of 100 autonomous "Worker Agents" that rotate to ensure zero downtime and lightning-fast data extraction.

​🛠 Key Features:
​For General Users:
🔹 Asset Intelligence: Instant pricing for Usernames, Gifts, and (soon) +888 Numbers + visual content generation.
🔹 Deep Wallet Tracking: Cross-referencing On-chain assets with active Telegram usage (verifying if a username is actually deployed on an active account).
​For me/Admin (Agent Tools):
🔹 Growth Engine: Autonomous agents that analyze Gift transactions to identify and onboard high-value users into the my channel.
🔹 Content Automation: An internal agent that monitors market posts, regenerates them using our branded templates, and writes professional captions instantly for our channels.
Running live in @FragmentsCommunity and serving our private whale investors group, @FragmentInvestors.

​Try it now:
🤖 Bot: @iFragmentBot
📢 Community: @FragmentsCommunity

​We are ready to scale with your support! 🚀

Hi everyone. I'm the founder of the @ton_ai_core project. I was one of the first to start the entire AI movement on TON. The project's goal was to teach AI to write smart contracts. A bunch of subprojects were created for this purpose. I rewrote Blueprint to make it easier to work with AI. I indexed all the contracts from the blockchain and created a bunch of other tools.

All ton chats: https://github.com/ton-ai-core/ton-chats/tree/main/out/markdown
Ton Api Indexing: https://github.com/ton-ai-core/ton-api-indexing
Dev template: https://github.com/ton-ai-core/contract-knowledge
Ton docs skill - https://t.me/ton_ai_core/30

I'm also currently developing a "new programming paradigm for AI coding."

I'm also creating a docker-git project to provide full agent access to the system. https://t.me/ProverCoderAI/29

I used this one as it has the most stars on Github. And it works well but I’m become worried of it using too much space in my context all the time

https://github.com/chigwell/telegram-mcp

But be careful: I once asked my Claude to remind me something and this MCP posted a message directly to my telegram channel. 💌

Instead of MCP I’d much prefer a skill + CLI tool as it saves the context.

Hey everyone,

🛡 CleanAds AI Agent
An autonomous escrow and brand-safety oracle for Telegram advertising. Advertisers lock TON in a smart contract; the off-chain AI Oracle monitors publisher channels in real-time. It analyzes context to prevent scams/profanity and triggers instant on-chain payouts via TON only if the content is AI-approved.

CleanAds AI Github
Live Demo TMA