Nic Silver

I see a lot of serious posts about AI. Not enough people sharing their weird experiments. We get so caught up in solving big problems we forget to just explore. Don't forget to test the limits and break stuff. Many of my own experiments have led me down interesting paths I wouldn't have found otherwise.

My AI agents kept failing until I fixed something most people skip. Right now I'm building a workspace where all my agents and workflows live in one place. It wasn't planned out from the start but it's something that keeps growing as I hit bottlenecks and solve them with AI. Think stuff like lead gen, content campaigns, prospecting etc. Whatever slows me down, I add it to the system so I can later hand it off completely. And as I keep working with it I'm refining it. Fixing what's not effective and tightening what is. But I've noticed something in this process. Having agents that do tasks for you is only one part of the equation. You NEED a strong foundation underneath it all. This means having a solid context layer and clear standard operating procedures. Context is the easier part. This means giving the system access to your: → past content → meeting transcripts → sales call notes → ICP & brand docs The stuff that you already have but sits scattered across different folders and software. Process documentation takes more effort. Here is where you need a real understanding of the operational layers of your business to map it out properly. Without those two things the agents are essentially guessing and having to fill in blanks. With them, the whole system starts to feel like it actually understands your business. It knows what you want, what you need and it only gets better over time. The hype bros want you to think you can build this overnight but you can't. It grows WITH your business and gets better the more you use it. This tech is still rough around the edges. But the people building their own agentic systems right now are going to be way ahead when the next wave hits. Seeing it come together for myself has me more excited about AI than I've been in a while. What does your current setup look like? One connected system or still piecing things together?

Agent memory has been a big headache for me... I've used Supabase, Pinecone, you name it. Probably tried them all. They store stuff just fine but when it comes to retrieving tool calling context and agent-specific info they fall short. Honcho is different. It's not a general purpose database you bolt onto your agent. It has built-in reasoning that decides what's actually relevant before it retrieves anything. Been testing it the last few days and the retrieval quality is noticeably better than my previous setups. The other big plus is portability. It works across OpenClaw, Claude Code, Hermes, all of it. This gives you one single memory layer without the migration headaches when you switch frameworks. Check it here: honcho.dev

@UrbanGibon I’m out… I saw brain required and that was enough. I pay for ai so I don’t have to think

Don't let the goofy drawing fool you... This is one of “those articles” you won't find anywhere else on X (not tooting my own horn, just telling the truth) So here's the deal: • Click the article below • Read for 10 sec • Decide A - Keep reading B - Leave Cool?

@AntonMartyniuk interesting, will check it out

AI agent forgot the implementation details we discussed yesterday And that's the real problem Every team building AI agents hits the same wall. A user talks to your agent on Monday. Gives it context, code examples, PR history. On Tuesday, the agent has zero memory. The user starts from scratch. Most teams try to fix this in 3 ways. All fall short. ❌ Expanding context windows Bigger context windows are just expensive sticky notes. Your conversation fits, but disappears the moment you refresh. No learning, no memory, no persistence. ❌ RAG systems can't connect the dots They retrieve documents but can't remember that YOU prefer specific architectures, or that your team decided against microservices last month. ❌ Most companies patch together 3+ databases Vector DB for embeddings, graph DB for relationships, SQL for metadata. Result: fragile architecture, security nightmares, and zero shared transactions. I spent weeks researching how to solve persistent agent memory without building a Frankenstein stack. That's when I found Oracle's AI Database approach. Instead of using multiple systems, ONE database handles everything: → Vectors for semantic understanding → Graphs for relationship mapping → Relational data for business context → All with ACID compliance across data types This means when your agent stores a new memory, vectors, graphs, and relational data update together. No partial writes. No inconsistent states. 📌 What stood out to me the most: Row-level isolation ensures your conversations stay private. Full EU AI Act and GDPR compliance, including "right to be forgotten" while maintaining 10-year audit trails. If you are building AI agents that need to remember users across sessions, this is worth exploring. 👉 Get started with the Oracle AI Database free resources: fandf.co/4rbGwb8 —— ♻️ Repost to help others fix AI agent memory ➕ Follow me ( @AntonMartyniuk ) to improve your .NET and Architecture Skills Many thanks to @oracle for sponsoring this post

@IAmAaronWill yeah just because something can be automated doesn’t mean it should be.

Things I would never automate with AI: > Comments > Sales calls > Positioning > Offer creation > Client relationships > Objection handling > High stake DMs/emails Automate the inputs. Keep the outputs human.

@itsolelehmann yup, there are often just a handful of tasks that truly moves the needle. its fun to build new systems and all that, but if you are not showing it to others or reaching out offering your services, nobody is going to know about it.

it's very easy to lose yourself in AI fake work because you can create a lot of outputs, you can trick your brain into thinking you are productive but your standard response should be NO to most tasks most time you - don't need more data (just another addiction) - don't neeed more research you need to do the things that you've been avoiding

@michlimlim Oh no!

Claude is down. I fear I may have to form an original thought.

@janetmachuka_ @copy_ai @WriteSonic @rytr_me @OpenAI @invideoOfficial @pictoryai @synthesiaIO @canva @surferseo @semrush Step 1 cancel ChatGPT subscription. Step 2 subscribe to Claude. Step 3 use Claude Code or Cowork to build whatever you need specific to your business.

To all marketers who are only stuck with ChatGPT, explore these AI marketing tools. 1. Content Creation: @JasperAI, @copy_ai, @writesonic, @rytr_me, @OpenAI 2. Video & Creative: @InVideoOfficial, @pictoryai, @synthesiaIO, @canva 3. SEO & Optimization: @SurferSEO, @semrush, @SERanking 4. Email Marketing & Lead Generation @ConvertKit, @GetResponse, @HubSpot, @Zoho 5. Website, Landing Pages & Funnels: @Leadpages 6. Social Media & Engagement: @semrush, @drift, @Zendesk, @freshworks 7. Automation & Workflow: @zapier 8. Analytics & Insightsl: @tableau, @DataRobot Which ones are you exploring at the moment?

@KrishXCodes agreed!

After building AI agents in production one thing is clear: The hardest part isn't the LLM. It's: > context retrieval > tool orchestration > memory > evaluation Prompt engineering was the easy part 🙃

@iamgalba Yes, learning to use Claude Code and build custom workflows is 10000x more valuable than chasing shortcuts.

There will be more of these “AI marketing killers”. 99.9% of that is basic wrapper that does no more than Manus. Filter the noise. Where AI is definitely is killing it - in well structured AI agentic flows built and executed by Claude Code based on YOUR processes. For YOUR brand. This is where 20-50x output multiplier happens. Not in nice generic dashboard with AI sticker slapped on it.

@adilmania yup there are no shortcuts to good marketing. AI tools can be helpful but they won’t be able to do all the work for you.

this going viral means 1 thing: dev will do everything they can to skip marketing what they built, thinking a tool will fix it by magic. marketing is all about intuition, iteration & taste. all AI marketing tools i've seen (at least for now) are slop & spam companies.

@shannholmberg Yup, there will never be a plug and play CMO AI solution. Needs to be tailored to each company and their specific needs.

everyone wants an AI CMO truth is there's no product out there yet that actually delivers on it. what you can do is build your own AI growth system using your data, your voice, and your workflows. I mapped out how to build it in the 5 levels of AI marketing framework.

@SamPeterToT AI services and AI automation services are top for me. There's a huge demand

Cashflow focus right now: ➣ web3/AI jobs ➣ AI services ➣ bounty hunting ➣ ambassador programs ➣ paid ghostwriting for X/LinkedIn ➣ AI automation for agencies ➣ BD for LinkedIn (founders/projects) ➣ UGC + AI video production Which one are you stacking first?

@0xmani @Polymarket Wow, that’s impressive! Need to look into that

This guy turned $1,500 into $7,429 on @Polymarket in a week without placing a single trade himself. 6 AI agents did it. $29/month total cost. running 24/7 Scanner - watches BTC/ETH markets, streams live prices from Binance Factor Miner - finds trading edges using Claude Haiku, kills underperformers automatically Analyst - combines ML model + news sentiment + whale orderflow into one probability score Auditor - blocks bad trades, flags fake news, kills low liquidity plays Risk Manager - quarter kelly sizing, shuts down after 3 losses in a row Executor - places orders automatically with retries 105 trades. 69 wins. 65.7% win rate.

@JakehellerAI Yup, in just the last few months AI has become dangerously good... but it still need human oversight. But we are making some major advances in the AI space right now. So only a matter of time before they can run on their own.

Pre covid: hire locally 2020: offshore to the Phillipeans 2026 and beyond: automate with AI The offshoring era will be very short lived

@RoxanaLimban Ohh, this is awesome! Will get on this asap

Anthropic just launched Anthropic Academy, a free AI learning platform with 13+ courses on Claude, APIs, MCP, and building AI agents. Tools are here Information is free Lock in.

@AutomationKing0 Doing what is already working is the single best advice I got in business. When I started I wished someone had told me. Would have saved me many failed product launches...

Crafting a product that actually sells requires a lot of research and competitor analysis. I have spent over 3 hours in the Meta Ads Library. Studying what AI automation specialists are running ads for. Their offers. Their hooks. Their funnels. The patterns are interesting. Time to cook something dangerous.

@rumit_s_anand The ability to self-anneal and learn from its mistakes to become better over time.

Most “AI agents” today are just chatbots with tools. But real agents need much more. Contest time 👇 In one line: What is the ONE thing an AI agent must have to truly act autonomously? Best answer wins an ₹500 Amazon voucher. #AI #AgenticAI #AIAgents #ArtificialIntelligence

@levie Good point! Not often talked about, but makes sense.

“AI exposed jobs may increase hiring and attract higher wages. It all depends on a) elasticity of consumer demand and b) number of AI exposed tasks in a job.” This is a key point. We’re going to see lots of AI automation emerge that has the opposite effect that we expect, because the cost of doing something goes down and greater demand for that service exists at lower prices. Take a *very* simplistic example in agentic coding to see what happens when you can dramatically increase output per $ of engineering budget. Before AI, a mid-sized company or team within a large company has a project they want to build software for. It takes 50 engineers to fully resource the effort, but the project doesn’t provide the ROI to fund it compared to other initiatives. Or the company knows its expertise isn’t in building software so it’s not even worth starting. So they hire 0 engineers, and don’t start the project. Now, AI agents make it possible for this to be a 10 engineer problem. All of a sudden the ROI calculus immediately changes on starting up the project. So now instead of hiring 0 engineers to do the project, the company hires 10 with AI agents. This has endless implications in coding, in particular, because coding can now have impact for anything from doing internal workflow automation, systems integration, data analysis, as well as customer-facing product innovation. By bringing down the cost of writing code, we can just begin to use it for far more. This will likely play out in a number of other job families as well, where lowered costs or higher output will lead to more demand. Now, not all of this will be smooth. For instance, there may need to be some reallocation of talent across the economy to move from some places of excess supply to places of lower supply. This could be bumpy at times, but the dynamic holds.