Tekkeys

01/12/2025

Let’s be honest. Keeping a social media presence alive while running a business is a nightmare.

We all know we need to post. We need to be consistent. We need to "build a brand."

But the reality? Drafting, editing, finding images, and scheduling takes hours. Hours that you should be spending on building your products.

That’s why we at Tekkeys are building Autopost.

It is not just another scheduler. It is an AI Agent. You give it one raw thought, a single sentence or a messy note. It turns that into a full, optimized campaign across LinkedIn, Facebook, Instagram, X and etc.

We are automating the "annoying" parts of growth so you can focus on the work.

We are opening the Waitlist for you. If you want to reclaim 10+ hours a week, get in line early.

👇 Join the Waitlist: https://autopost.tekkeys.com/

13/11/2025

In the rapidly evolving field of artificial intelligence, large language models (LLMs) have impressed us with their ability to write, chat, and summarize. However, they have traditionally operated in a "closed world." Ask one a question, and it generates an answer based solely on the data it was trained on, without the ability to check new facts or perform real-world tasks. This can lead to "fact hallucination" or an inability to solve multi-step problems.

Enter the ReAct framework, a groundbreaking approach that gives LLMs a new set of capabilities. ReAct, which stands for "Reasoning and Acting," transforms a passive LLM into an active "agent." It's a system inspired by the way humans solve problems: we think, we act, we observe the result, and then we think again.

This framework was introduced by Yao et al. in 2022 to allow LLMs to generate not just text, but also reasoning traces and task-specific actions in an interleaved manner. This synergy allows the AI to create and update action plans, handle exceptions, and, most importantly, interact with the outside world.

How It Works: The Thought-Action-Observation Loop
At the heart of the ReAct framework is a simple but powerful loop. Instead of just outputting a final answer, the agent repeats a three-step process:

Thought: The agent first reasons about the task. It analyzes the user's request, breaks it down into smaller steps, and forms a plan. This is an internal monologue, like, "The user is asking a complex question. I first need to find piece of information A."
Action: Based on its thought, the agent then chooses an action to perform. This isn't just writing text; it's deciding to use an external tool. This could be Search[price of gold] or Calculator[98000 * 150000].
Observation: The agent receives the observation from its action, the search result, the answer from the calculator, or an error message.
This observation then feeds into the next thought. The agent might think, "Okay, I found piece A. Now I need to use this information to find piece B." This Thought-Action-Observation loop continues until the agent has gathered all the information it needs to formulate a complete and accurate final answer.

The Core Components of a ReAct Agent
To function, a ReAct agent relies on several key components working in concert:

1. The "Brain" (The LLM): This is a powerful language model that provides the core planning and reasoning capabilities. It's the "thought" part of the loop.
2. The "Hands" (External Tools): This is a "toolbox" of utilities the agent can use. It can include anything from a Calculator for precise math to a Search API for accessing real-time information, overcoming tasks LLMs traditionally struggle with.
3. The "Rulebook" (The Agent Framework): This is the core logic that connects the LLM to the tools. Often implemented using a library (like LangChain), this framework is the "operating system" that tells the agent how to follow the ReAct loop and decide which action to take.
4. The "Notebook" (Memory): To be effective, an agent needs to remember what it has already done. A memory component stores the history of thoughts, actions, and observations, allowing the agent to refer to previous steps and avoid repeating work.
Why This Matters: From Chatbot to "Do-bot"
The ReAct framework is a significant leap forward for AI for two main reasons:

Knowledge-Intensive Tasks: For complex questions that require piecing together multiple facts (like those on the HotPotQA benchmark), ReAct agents can dynamically search for, retrieve, and synthesize information, leading to more reliable and factual answers.
Decision-Making Tasks: In interactive environments, like text-based games (ALFWorld) or simulated shopping websites (WebShop), ReAct agents can navigate complex decision trees, explore their options, and act to achieve a goal.
By combining the reasoning power of LLMs with the ability to act and gather new information, ReAct agents become more trustworthy, interpretable (we can read their "thoughts"), and capable. They are a foundational step away from simple chatbots and toward the complex, "agentic systems" that can understand a goal, make a plan, and execute it in the real world.

12/11/2025

Google's Pomelli Didn't Destroy Autopost. It Proved Why You Need It.

A quick spoiler: No. In fact, it just proved exactly why you need a true automation platform more than ever.

You’ve probably seen the headlines. Google just dropped Pomelli, a new AI experiment from Google Labs that generates on-brand creative assets for businesses.

You give it your website, and it scans your "Business DNA", your colors, fonts, and tone to instantly create professional-looking images and ad copy. It’s slick, it’s impressive, and it’s from Google.

As the co-founders of Tekkeys, we’re building Autopost, an all-in-one social media automation platform.

So when a giant like Google steps into our space, we have a "gulp" moment. Did they just build a tool that makes ours obsolete?

We spent time with Pomelli, and our "gulp" moment quickly turned into a moment of clarity. Pomelli is a fantastic feature, but it’s not a solution.

It doesn't destroy Autopost. It just shines a massive spotlight on the one problem it doesn't solve.

The "So What?" Problem: The Gap Pomelli Leaves Behind
Let's walk through the user's journey with Pomelli.

You go to Pomelli.
You enter your website URL.
You get a set of beautiful, perfectly on-brand images and captions.
You download them to your computer...Now what?

This is the gap. This is the real work. You're left with a folder of assets and the same manual, time-consuming job you had before:

You still have to manually log in to LinkedIn, Facebook, Instagram, X, and every other platform.
You still have to figure out the best time to post for each audience.
You still have to set a reminder to post it at that exact time.
You still have to track the performance by logging back into every single platform to see what worked.
You still have to repeat this entire manual process tomorrow, and the day after, and the day after that.
Pomelli is like an AI-powered chef that hands you a box of perfectly chopped, high-quality ingredients. But you're still the one who has to cook the meal, serve it, and wash all the dishes.

Autopost Isn't an Ingredient. It's the Entire Automated Kitchen.
This is the fundamental difference. Autopost wasn't built to just make content; it was built to automate the entire workflow.

It's the difference between an asset generator and an automation engine.

Pomelli solves a visual consistency problem. Autopost solves a workflow and time problem.

A Feature is Not a Platform
We're excited about tools like Pomelli. They push the industry forward and democratize high-quality design, which is great for everyone.

But we're not worried.

We know that for busy founders, marketers, and creators, the real bottleneck isn't just thinking of what to post. It's the entire, relentless, manual process of scheduling, publishing, and tracking, day in and day out.

That's the problem we solve.

So, no, Google Pomelli didn't destroy Autopost. It just made our mission crystal clear.

Google built a great feature. We built the complete solution.

Want to stop juggling tabs and get your time back?

Autopost automates your entire social media workflow from idea to publishing and performance.

10/11/2025

If you run a Small or Medium-sized Enterprise (SME), your to-do list is endless. You’re the CEO, the head of sales, the accountant, and often, the reluctant social media manager.

You know you need to be on social media. It’s where your customers are. But finding the time to do it and do it well feels impossible.

You post consistently for one week, get swamped with a big order, and your social media channels go silent for a month. This inconsistency doesn't just slow your growth; it can make your business look unprofessional or, even worse, closed.

What if you could get all the benefits of a professional, consistent social media presence without spending hours on it every day? What if you could put the entire process on autopilot?

This is the exact problem Autopost was built to solve. It’s not just another scheduling tool; it’s an end-to-end AI platform designed to take you from a single idea to a fully published campaign.

Here’s how Autopost will become a growth engine for your SME.

1. It Cures "Blank Page Syndrome" with AI Strategy
The Problem: You sit down to create a post and... nothing. You don't know what to post, what hashtags to use, or what your customers even want to see.

The Autopost Solution: Autopost’s AI does the strategic work for you. It doesn't just offer generic ideas. It analyzes current trends in your specific industry to suggest topics, captions, and hooks that are performing right now. It hands you a ready-made content plan, turning a 60-minute brainstorming session into a 60-second decision.

2. It’s Your 24/7 AI Copywriter and Designer
The Problem: You’re not a graphic designer or a professional copywriter. Your posts can look amateurish, with stretched logos or captions that don't sound right. This damages your brand.

The Autopost Solution: Autopost generates content in your brand's unique voice. It writes compelling captions and can generate visuals that look professional and consistent. For an SME, this is like having an in-house marketing team without the massive overhead. It levels the playing field, allowing you to compete with the look and feel of a much larger company.

3. It Buys You Back Your Most Valuable Asset: Time
The Problem: Manually logging in and posting to Facebook, then Instagram, then LinkedIn is a time-consuming, soul-crushing chore. This is time you should be spending talking to customers or developing new products.

The Autopost Solution: This is the core of our "Automating the Annoying" mission. Autopost’s smart-scheduling calendar lets you plan your entire week or month of content in one sitting. It then automatically publishes everything at the optimal time for each platform. You are literally building your brand and generating leads while you sleep or while you're busy running your actual business.

4. It Replaces Guesswork with Growth Data
The Problem: You’re posting into the void. You have no idea if your efforts are actually working, driving sales, or reaching new people.

The Autopost Solution: Autopost closes the loop with clear, simple performance insights. You’ll see exactly which posts are driving engagement, what content your audience loves, and where your growth is coming from. This allows you to stop wasting time on what doesn't work and double down on what does, turning your social media from a chore into a predictable growth channel.

A Bonus for Sri Lankan SMEs: A Tool That Speaks Your Language
Here is the Tekkeys difference. Most global social media tools are built for a Western audience. They don't understand local context, culture, or nuance.

Autopost is being built by a Sri Lankan team with a "Sri Lanka-first" principle. Our long-term vision is to create a powerful AI that can deeply understand and generate content in Sinhala and Tamil.

Imagine an AI that can't just post in your local language but can understand the cultural context, local holidays, and trends that matter to your customers. This is the future we are building, a tool that gives our local businesses a true competitive edge.

Conclusion: Focus on What Matters
Autopost isn't just a social media tool; it's a business tool. It’s designed to free you, the SME owner, from the role of "content creator" so you can get back to being a "business builder."

Stop just posting. It's time to start growing.

Ready to put your social media on autopilot?
https://autopost.tekkeys.com

01/11/2025

We was just honored as one of Google Cloud's biggest supporters! 🎉

01/11/2025

With Google – We got recognized as one of their top fans! 🎉

01/11/2025

With Google for Startups – We just got recognized as one of their top fans! 🎉

01/11/2025

𝐓𝐢𝐫𝐞𝐝 𝐨𝐟 𝐠𝐞𝐧𝐞𝐫𝐢𝐜 𝐀𝐈 𝐜𝐨𝐧𝐭𝐞𝐧𝐭 𝐭𝐡𝐚𝐭 𝐬𝐭𝐢𝐥𝐥 𝐧𝐞𝐞𝐝𝐬 𝐡𝐨𝐮𝐫𝐬 𝐨𝐟 𝐦𝐚𝐧𝐮𝐚𝐥 𝐰𝐨𝐫𝐤? 𝐖𝐞 𝐚𝐫𝐞 𝐛𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐭𝐡𝐞 𝐬𝐨𝐥𝐮𝐭𝐢𝐨𝐧.

The problem with most AI tools is they're "monolithic" They're a single consultant. We wanted to build an entire department. That's why we built Autopost on the Agent Development Kit (ADK), a powerful multi-agent framework. This allows us to orchestrate a full team of AI specialists that collaborate just like a human content team.

Read the full article: https://autopost.tekkeys.com/chronicles/the-automation-paradox-adk-vs-genkit

Written by Inusha Gunasekara

A guide for tech leaders on choosing the right AI orchestration framework. Compare ADK's multi-agent system, GenKit's app-first approach, and LangChain.

01/11/2025

𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐨𝐧:

As AI development matures, the conversation has shifted from simply using models to orchestrating them. This has led to a confusing landscape of frameworks, each with a different philosophy. While tools like LangChain pioneered the orchestration space, a new generation of Google-backed frameworks, namely GenKit and the Agent Development Kit (ADK), have entered the field.

However, a direct comparison is often flawed. These tools are not just different-flavored solutions to the same problem; they are designed to solve fundamentally different types of problems. Choosing the right one is less about which is "best" and more about matching the tool to your architectural needs.

𝐔𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐭𝐡𝐞 𝐏𝐥𝐚𝐲𝐞𝐫𝐬: 𝐓𝐡𝐫𝐞𝐞 𝐃𝐢𝐬𝐭𝐢𝐧𝐜𝐭 𝐏𝐚𝐫𝐚𝐝𝐢𝐠𝐦𝐬

𝟏. 𝐋𝐚𝐧𝐠𝐂𝐡𝐚𝐢𝐧: 𝐓𝐡𝐞 𝐒𝐢𝐧𝐠𝐥𝐞-𝐀𝐠𝐞𝐧𝐭 𝐎𝐫𝐜𝐡𝐞𝐬𝐭𝐫𝐚𝐭𝐢𝐨𝐧 𝐓𝐨𝐨𝐥𝐤𝐢𝐭
LangChain was the first to popularize "chaining" the idea of gluing LLM calls together with tools, data sources (like vector stores), and memory. It provides a flexible, low-level, and extensive toolkit for building a single, complex workflow. Think of it as a set of building blocks for one agent to perform a sophisticated, multi-step task.

𝟐. 𝐆𝐞𝐧𝐊𝐢𝐭: 𝐓𝐡𝐞 𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧-𝐅𝐢𝐫𝐬𝐭 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤
GenKit is a framework for embedding generative AI features into an existing application. Built by the Firebase team, it is "app-developer-first," prioritizing observability (traces, logs), reliability, and a clean deployment path. It is less concerned with complex agentic reasoning and more with providing a production-ready "flow" like a RAG endpoint or a summarizer that your application can call. It is available for Node.js and Go.

𝟑. 𝐀𝐃𝐊 (𝐀𝐠𝐞𝐧𝐭 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐊𝐢𝐭): 𝐓𝐡𝐞 𝐌𝐮𝐥𝐭𝐢-𝐀𝐠𝐞𝐧𝐭 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤
This is the most misunderstood of the three. ADK is not just another "AI kit"; it is an Agent Development Kit. Its entire purpose is to build, deploy, and manage multi-agent systems. The ADK paradigm is not a single chain, but a team of specialized agents that collaborate, delegate, and communicate (using Agent-to-Agent, or A2A, protocols) to solve problems that are too large or complex for any single agent.

𝐓𝐡𝐞 𝐂𝐨𝐫𝐞 𝐃𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐜𝐞: 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐚𝐥 𝐀𝐩𝐩𝐫𝐨𝐚𝐜𝐡

The confusion between these tools arises from viewing them all as "orchestrators." A better way to understand them is through an analogy:

LangChain is like hiring a single, hyper-capable consultant. You give them a set of tools (a phone, a database, a calculator) and a complex set of instructions (a chain) to follow. They do the entire job themselves, from start to finish.
GenKit is like installing a new software suite (ex:- Salesforce) in your company. It's a production-ready, observable feature that your existing application can now use to perform a new, specific function.
ADK is like building an entire department. You hire a "manager" agent who coordinates a team of specialized agents: a "researcher," a "writer," and a "fact-checker." They delegate tasks and pass work between each other to complete a large-scale project.

LangChain and GenKit offer control vs. abstraction for features. ADK offers a new paradigm for collaborative systems.

𝐔𝐬𝐞 𝐂𝐚𝐬𝐞 𝐒𝐜𝐞𝐧𝐚𝐫𝐢𝐨𝐬: 𝐖𝐡𝐞𝐧 𝐭𝐨 𝐂𝐡𝐨𝐨𝐬𝐞 𝐖𝐡𝐚𝐭

Choose LangChain if: You are building a single, complex agent (ex:- a ReAct agent).You need to rapidly prototype a custom RAG (Retrieval-Augmented Generation) pipeline. You want fine-grained control over every step of the orchestration logic. You need to leverage the largest ecosystem of third-party tool integrations.
Choose GenKit if: You are an application developer. Your primary goal is to add a production-grade AI feature (like a chatbot or summarizer) to an existing application. Production-grade observability (tracing, logging, monitoring) is your main priority.
Choose ADK if: Your problem is too complex for one "monolithic" agent. You need to orchestrate a team of specialized agents that must collaborate. Your use case involves delegation (ex:- a "planning" agent that dispatches tasks to "ex*****on" agents).You are building a system where different agents (potentially built by different teams) need to communicate.

𝐂𝐨𝐧𝐜𝐥𝐮𝐬𝐢𝐨𝐧: 𝐀 𝐍𝐞𝐰 𝐓𝐨𝐨𝐥𝐤𝐢𝐭, 𝐍𝐨𝐭 𝐚 𝐑𝐞𝐩𝐥𝐚𝐜𝐞𝐦𝐞𝐧𝐭

The arrival of ADK and GenKit does not eliminate the need for LangChain. Instead, these frameworks provide more specialized tools. Developers can even "hybridize" for example, by using a LangChain-built chain as a specific "tool" that one of an ADK's agents can use.

The best approach depends on your architectural philosophy. If you need a flexible toolkit for a single agent, use LangChain. If you need to ship an observable AI feature in your app, use GenKit. And if you are ready to build a system of collaborating specialists, use ADK.

“In the rapidly evolving space of AI tooling, developers are now liberated to choose not only what they can build, but in fact the very manner in which they can orchestrate intelligence itself.”

This choice is at the heart of our work at Tekkeys, where we're using the ADK's powerful multi-agent framework to build AutoPost, our next-generation AI automation tool.

-Inusha Thathsara

17/10/2025

HUGE NEWS! We've Joined the Google Family! 🤯

​We are thrilled to announce that Tekkeys has been accepted into the Google for Startups Cloud Program!

​This is a massive win for local entrepreneurs. It means we have the essential Google Cloud credits and technical access to build and scale AutoPost into the most reliable content automation system available.
​What does this mean for your business?
​It guarantees that our AI Agent can handle complex tasks and deliver ultra-reliable service. We're committed to making your content creation effortless and accurate, especially in Sinhala and Tamil.
​This partnership accelerates our mission to end the content crisis for SMEs.

​Ready to stop wasting hours on social media?
Join the early access program today:

autopost.tekkeys.com

​