Berkay M

We benchmarked Browser DevTools MCP vs Playwright MCP on the same flow: → 78% fewer tokens → 76% fewer turns → 66% lower cost → 57% faster → Much more consistent run-to-run How we did it (tool design, execute batching, scoping): @serkan_ozal/browser-devtools-mcp-78-fewer-tokens-vs-playwright-mcp-faster-and-more-consistent-32f314004d30" target="_blank" rel="nofollow noopener">medium.com/@serkan_ozal/b…

I read the ClickHouse blog post (clickhouse.com/blog/llm-obser…) where an AI assistant uses MCP to query OpenTelemetry data in ClickHouse and perform root cause analysis. It works to a point. Traces, metrics, and logs are great at localizing a problem. But in real production incidents, localization is not always the same thing as understanding the actual root cause. Once you know where the problem is, sometimes you need to go deeper. Application state. Runtime behavior. Host, VM, container, or Kubernetes level signals. That data usually isn’t collected continuously because it’s expensive and adds overhead. This is exactly the problem we are working on. Instead of collecting everything all the time, we use AI to decide when a problem happens and then collect deep, high fidelity data on demand across the full stack. Application, infrastructure, and environment. Safely, with guardrails and PII controls. The ClickHouse experiment shows what AI plus observability can already do. The next step is giving AI the ability to actively investigate systems when it matters, not just explain dashboards. That is where full stack autonomous debugging starts. #observability #opentelemetry #AI #debugging #SRE #production

@faik There is also shift from doing all the work and "giving" the context to the LLM to, designing the LLM, giving the right instructions and tools, so that LLM can interactively construct its context. Instead of deterministic workflows we implement, let the LLM control the flow.

There’s a growing shift in the AI space from Prompt Engineering to something more practical and impactful: Context Engineering. Prompt engineering focuses on crafting a single input string. But with modern AI Agents, that is no longer enough. These systems need much more than a clever prompt to function reliably. Context Engineering is about assembling all the information an AI needs to perform a task. That includes the user’s current request, system instructions, memory of past interactions, retrieved documents, relevant tools, and structured output formats. Most failures in agents today are not model problems. They are context problems. The agent did not have access to the full picture, and so it made the wrong decision. Designing good context is not just a technical task. It requires understanding the user, the use case, the workflows, and how to present the right information at the right time. As AI systems get more capable, Context Engineering will be the foundation for building agents that actually work. One of the challenges is that context is dynamic. What matters changes from task to task. Scheduling a meeting might require access to your calendar, recent emails, and tone preferences. Writing a report might require historical project data and a specific output format. The system needs to decide what is relevant, fetch it, structure it, and feed it to the model, all before the model is ever called. This also means that Context Engineering sits at the intersection of system design, retrieval, memory, tooling, and product thinking. It is not just about giving more data. It is about giving the right data, cleanly presented, and only when it helps. We are moving from writing better prompts to building better environments around the model. The most reliable agents are not those with clever instructions but those with well-designed, thoughtful context flows. #AI #ContextEngineering #LLM #Agents #AIInfrastructure #SystemDesign #AIProducts

@Kdrnp @polymetai useful, thanks for the resources! Q: assuming your potential users are non-designers, is it possible for @polymetai to have this knowledge & guide me, with choices/examples, teach me in gradually in product rather than me learning how to talk its language elsewhere?

@polymetai If you’re a PM experimenting with AI-driven design, spending an hour with the resources above will save you many more hours in prompt tweaking and review cycles. Always happy to trade notes with anyone exploring similar ground.

We noticed at @polymetai, one pattern keeps surfacing: the output quality rises or falls with how well the prompt speaks the language of UI & UX. You don’t need to turn into a designer, but a working grasp of three fundamentals makes totally huge difference:

@irensaltali Good stuff @irensaltali! If you're in this field, no matter the role, must be exploring these tools, new workflows, etc.

🚀 Just Launched: My own Micro SaaS app powered entirely by Open Source & AI! 🤖✨ 🔍 Curious how I built it with minimal resources? Here’s the breakdown… 👇 🍍 It's like a fun project but I'll keep maintaining. #SaaS #OpenSource #AI #MicroSaaS

@rauchg @v0 lol awesome, I can't keep up :)

@berkay @v0 x.com/v0/status/1843…

It’s addictive to see @v0 ship an instant change. Fast is fun.

@serkan_ozal may be use @clerk?

I am off today (Labor day) and playing with React for a login/signup page after soooo many years of backend oriented career. AWS Amplify really makes it so easy. But I am still open to other alternatives if you think there are easier ones :)

@adamhjk One of the problems that we're having with the meetings, in person or video, is that there is no record of what's discussed that others can see. How are you handling this? Any use of tools to summarize meetings, etc.?

We build the whole company around using slack to facilitate video. It’s fantastic.

@adamhjk @pseudomata I read the suggestion in the doc as more communication not less. Don't say just hi (signaling you want to chat), but say hi and provide context to enable the other side to decide when they should engage depending on their mood, availability etc.

@pseudomata Yay! But that’s what I mean - the solution here isn’t less communication, or different communication - it’s more communication! Fewer rules, more talking. :)

You know how our slack starts every morning? “Good morning!” ☀️ it’s a race to see who says it first (the Europeans usually win). Slack and zoom facilitate real conversations, and real conversations involve pleasantries. Can you imagine a no hello rule in meat space?!

After working hard with our team, today is one of the biggest days of my life. We are proud to launch @peakacom on Product Hunt. We love to get your support. producthunt.com/posts/peaka youtube.com/watch?v=uR1d2p…

@irensaltali Nice work @irensaltali I love the idea! Even as a non-developer, I was able to get it up and running.

I have released Serverless API Gateway v1.2.0, which introduces a flexible and powerful method to dynamically manipulate API requests and responses using mapping configurations. #serverless #apigateway

I really don't like the #GitHub actions interface when it comes to running an action 😡 In a few minutes, I created a workflow to run my deploys in only one click. 🎥Watch how I did it: youtu.be/8TD2W2pAMr8 #Slack Check the full story: actioner.com/blog/deploy-co…

@gorkemcetin @actionercom

Türkiye’de B2B / B2C ürününe yapay zeka ekleyen firmaları arıyorum ama iki elin parmaklarını geçmiyor. Burada herkesin faydalanacağı bir liste yapalım mı?

As @Altogic team we were busy for the last few months in developing a new #Kubernetes based #opensource application development platform @AgnostDev. Let us know your comments and be part of the Agnost community. agnost.dev

Our browser extensions at Resmo (Safari, Chrome, Firefox, Edge, Arc, etc.) can catch weak passwords and even warn users on the fly about them before they create a password that goes against the company policy. If people overlook that warning, we can still catch it, and you can run remediation workflows through our Slack app, asking people to take action. 👊🏻 Weak passwords may seem obvious, but they remain one of the biggest reasons businesses get hacked. Resmo helps you monitor potential identity threats while keeping the privacy of your users in mind—no password is ever retrieved by us. If you want to learn more, ping me!

Thanks to @actionercom and #peerFeedback app, this is how easy it is to show gratitude. Using AI while writing the feedback is an always-wanted feature. BTW thank you @aysegulyg, as seen in the video `You are a great team worker`. 👏

Kudos app quickly became the most popular internal app @actionercom. It started as an hackathon project, great to see fruits of the the no-code platform.

@dgn which book? :)

Now this is an interesting start for a book 📖

Are you struggling with the time-consuming, back-and-forth process of finding someone to cover your on-call? Simplify it with Actioner 🚀 Manage everything on Slack and keep your entire team in the loop! Take a look at Actioner's on-call app for PagerDuty: tinyurl.com/bdft95w5