Konstantin Assanov

Make Claude Code and other coding agents 10x more powerful! Cipher is an open-source memory layer specifically designed for coding agents. Without memory, coding agents start every request without context. Cipher provides a persistent layer to store and share context so agents and the entire dev teams never lose track of project history and past fixes. Here's why it's a game changer: 🔌 MCP integration with any IDE you want. 🧠 Auto-generate AI coding memories that scale with your codebase. 🔄 Switch seamlessly between IDEs without losing memory and context. 🤝 Easily share coding memories across your dev team in real time. 🧬 Dual Memory Layer that captures System 1 (Programming Concepts & Business Logic & Past Interaction) and System 2 (reasoning steps of the model when generating code). ⚙️ Install on your IDE with zero configuration needed. It’s 100% open source. Link to the Github repo in the comments! If you're into ML, LLMs, RAG, and AI Agents, I share AI apps and tutorials every week. Subscribe to AI Engineering (it's free): https://lnkd.in/gfkzKZYk

1.623

68 Kommentare

AWS released the GitHub Actions for Lambda, but most of you shouldn't use it. It's ONLY useful for people who are managing their Lambda functions manually. If you're doing IaC already, then you shouldn't use this. Official doc here: https://lnkd.in/eBzi52RR This new GitHub Action lets you ship deployment artefacts to a Lambda function, and configure its settings (memory, timeout, etc.). But not the event triggers for the function, its IAM role, or anything else that makes up your application - API Gateway, Lambda permissions, DynamoDB tables, etc. If you're doing IaC already, whether it's with CloudFormation, Terraform, CDK, Serverless Framework, SAM, SST, Pulumi or any other tools, then you can already run deploy in a GitHub Action, e.g. (Serverless Framework) - name: restore NPM dependencies run: npm ci - name: deploy to dev run: npx sls deploy (CDK) - name: install CDK run: npm install -g aws-cdk - name: deploy to dev run: cdk deploy --all --require-approval never (Terraform) - name: terraform init run: terraform init -backend-config=config/dev.backend.hcl - name: deploy to dev run: terraform apply -auto-approve -var "stage_name=dev" See, it's not hard! Your entire application is deployed (and rolled back, if not using Terraform) as a single unit. And you run the same commands as you would in your terminal, so it's easier to debug. IF, you're not doing IaC today, and you're managing your AWS environment, including Lambda functions, manually. Then this new GitHub Action might be a useful baby step forward. However, you should really aim towards a proper IaC solution instead, one that can encapsulate your entire application, not just the Lambda functions in isolation.

465

51 Kommentare

I put together a step-by-step guide to help you implement LLM Optimisation the right way for E-commerce. In this carousel, you’ll find the high-level overview. The full version includes: → How to generate a structured product feed (CSV, JSON, XML) with the right attributes: title, GTIN, MPN, price, availability, and product URLs → How to apply Schema . org Product markup using JSON-LD, with full code examples → How to configure your robots.txt to allow GPTBot, OAI-SearchBot and PerplexityBot access to your PDPs and feeds → How to prerender JS-heavy product pages using static snapshots or SSR → Where and how to submit your feed to Perplexity’s Merchant Program and OpenAI’s early access initiative → How to debug AI visibility using server logs, schema validators, and live prompt testing inside LLM tools If you’re working on e-commerce, SEO, or growth, this is where things are moving. 💬 Comment or DM me, and I’ll send you the full guide. 📌 Save this post if you're not ready yet but know you'll need it.

580

914 Kommentare

The Great AGI Debate: Here's An Objective Breakdown Of Both Sides The question: Can Current LLM Architectures Ever Achieve Superintelligence? On one side, those who believe today’s large language model (LLM) architecture can manifest AGI with enough scale. On the other, those who believe only a fundamental architectural shift will get us there and the current LLM race is futile Side 1: Current Architecture Could Manifest AGI Belief: Scaling today’s transformer-based LLMs—more parameters, more compute, more data—plus incremental improvements (memory, multi-modality, better training) will close the gap to AGI. Key points: Scaling laws predict consistent performance gains as models grow in size and training data.¹ LLMs already show emergent abilities like in-context learning, reasoning chains, and professional-level exam performance.² Some reasoning-optimized LLMs now score above 85% on complex benchmarks like AIME and GPQA Diamond, rivaling or surpassing top human performers.³ Rapid hardware advances—such as projected millionfold increases in AI compute over the next decade—make massive scaling plausible.⁴ Leading voices at scaling-first companies have suggested AGI could emerge in the near term, possibly within a year.⁵ Side 2: Architectural Shift Required Belief: Current LLMs are advanced next-token predictors with no grounded understanding. AGI will require new cognitive stacks—world models, symbolic reasoning, embodied agents, and real-time learning. Key points: Large models still fail on ambiguous or novel problems, revealing brittle reasoning.⁶ Inverse scaling effects show that simply making models larger can degrade performance on certain tasks.⁷ Diminishing returns from scaling suggest limits without new approaches to representation, grounding, and planning.⁸ Some research groups envision AGI emerging from hybrid systems that integrate perception, reasoning, and action—not just text prediction.⁹ Expert surveys remain split, with many placing AGI decades away under current paradigms.¹⁰ Why This Matters If Side 1 is correct: The frontrunners are the scale champions with the most compute, data, and capital. If Side 2 is correct: The winners will be the innovators pioneering entirely new architectures. This isn’t just a technical debate. It’s a fork in the road for the entire industry—and it determines whether the future arrives through more of the same or a radical redesign. ******************************************************************************** The trick with technology is to avoid spreading darkness at the speed of light I’m the Founder & CEO of Curiouser.AI. We’re building Generative AI products to augment your ability to think, to elevate you and help you define your own limits. Our beta is named Alice, give her a try visit curiouserdotai or DM me

56

45 Kommentare

Selecting the Right Agentic AI Framework — Your Success Depends on It When building with Agentic AI, your choice of framework isn’t just a technical decision — it defines your product’s capabilities, scalability, and security. Here’s a structured breakdown of leading options and when to use them: 𝗔𝗗𝗞 (𝗚𝗼𝗼𝗴𝗹𝗲)  • Features: Modular AI agents, Gemini support  • Advantages: Flexible orchestration, rich tool ecosystem  • Best For: Conversational AI, complex autonomous systems 𝗟𝗮𝗻𝗴𝗚𝗿𝗮𝗽𝗵  • Features: Stateful workflows, graph-based execution, human-in-the-loop  • Advantages: Dynamic, traceable, adaptable  • Best For: Interactive storytelling, decision-making systems 𝗖𝗿𝗲𝘄𝗔𝗜  • Features: Role-based agents, dynamic task planning, conflict resolution  • Advantages: Scalable AI teams, collaborative intelligence  • Best For: Business strategy, project simulations, healthcare coordination 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰 𝗞𝗲𝗿𝗻𝗲𝗹  • Features: AI SDK integration, memory & embeddings, enterprise security  • Advantages: Enterprise-grade, scalable  • Best For: Workflow automation, enterprise applications 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝗔𝘂𝘁𝗼𝗚𝗲𝗻  • Features: Multi-agent conversations, custom roles  • Advantages: Simplified orchestration, robust error handling  • Best For: AI research, advanced chatbots, task planning 𝗦𝗺𝗼𝗹𝗔𝗴𝗲𝗻𝘁𝘀  • Features: Lightweight, modular  • Advantages: Low compute cost, fast integration  • Best For: Research assistants, small AI workflows 𝗔𝘂𝘁𝗼𝗚𝗣𝗧  • Features: Goal-oriented execution, adaptive learning  • Advantages: Self-improving, minimal human oversight  • Best For: Content creation, predictive analytics, automation The most important question isn’t “Which is the most powerful?” It’s “Which best matches my project’s complexity, scale, and goals?” 𝗔𝗰𝘁𝗶𝗼𝗻 𝗜𝘁𝗲𝗺𝘀 𝗳𝗼𝗿 𝗬𝗼𝘂  • Map your requirements — scale, security, human-in-the-loop needs.  • Prototype with one or two frameworks before committing.  • Choose something that can evolve with your future roadmap. 𝗬𝗼𝘂𝗿 𝗧𝘂𝗿𝗻  1. Which of these frameworks have you used or plan to explore?  2. Are there any emerging Agentic AI frameworks that should be added to this list?

1.106

64 Kommentare

When developers use "tech debt" to refer to any code that they didn't write (and some they did), how do you decide where to invest? I find this 2x2 quite helpful for framing the discussion. In many situations, the trouble with debt is not the principal, it's the interest. When it comes to technical debt, that's the increase in the cost of change that's created by the debt. And you're not really required to pay until you're working in that area. The axis for "ease of change" includes everything it takes to get a change into production. Is the code easy to reason about? Do you trust the tests? Is the pipeline fast? Anything that makes it harder is bad. However, if the system or component isn't undergoing change, it's not hampering your business as badly. Hence the need to incorporate "desired rate of change." In other words, if we could, how much work would we be doing in this area? If something is slowing you down or blocking you in an area of your system that presents critical opportunities to your business, it's probably time to get it cleaned up. Like any 2x2, it's pretty simple. But I assure you someone, somewhere, is arguing right now about how to improve a thing that's already good and pretty stable. PS. I used this chart in a workshop at LeadDev's LDX3 and it led to some great discussions. If you'll be at LeadDev in NYC or Berlin this year, come have a chat. The workshop will be repeated at both.

61

2 Kommentare