[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

@koylanai Avatar @koylanai Muratcan Koylan

Muratcan Koylan posts on X about ai, anthropic, open ai, context engineering the most. They currently have XXXXX followers and XXX posts still getting attention that total XXXXXX engagements in the last XX hours.

Engagements: XXXXXX #

Engagements Line Chart

Mentions: XX #

Mentions Line Chart

Followers: XXXXX #

Followers Line Chart

CreatorRank: XXXXXXX #

CreatorRank Line Chart

Social Influence

Social category influence technology brands stocks vc firms celebrities

Social topic influence ai, anthropic #62, open ai #1224, context engineering, llm #22, agentic, flow, prompt engineering #5, shell, $googl

Top accounts mentioned or mentioned by @openai @principalsai @elonmusk @shaanvp @anthropicai @tsarnick @salesforce @opencampusxyz @officiallogank @grok @xai @langchainai @sama @a16z @buildspace @andrewyng @huggingface @benioff @digicalidesign @teknium1

Top assets mentioned Alphabet Inc Class A (GOOGL)

Top Social Posts

Top posts by engagements in the last XX hours

"Even people who work in AI are missing that prompt design is literally everything. Cursor just published a blog post last week on how they updated their entire agent harness and prompt design specifically for OpenAI's new Codex model. - They renamed tools to match shell equivalents because the model responds better to familiar patterns from training - Simply providing a tool definition wasn't enough so they had to add explicit instructions like "After substantive edits use the read_lints tool to check recently edited files" - One prompt tweak telling the model to "preserve tokens" made it"
X Link 2025-12-07T17:27Z 9889 followers, 70.7K engagements

"OpenAI wants markdown structure. Anthropic prefers XML tags. Google emphasizes few-shot examples. So I built a simple agent system that reads the official prompting docs and applies them to the given prompt. Each optimizer runs a ReAct loop: - list_provider_docs discover available guidelines - read_provider_doc fetch specific content - submit_optimization return structured result Agent decides what context it needs. The loop runs until submission: X. LLM reasons about the task X. Calls a tool X. Observes result X. Repeats Typically 3-4 iterations. Parallel execution via asyncio.gather; all"
X Link 2025-12-05T11:31Z 9889 followers, 62.9K engagements

"Snipd has become my only way of listening to podcasts. I now spend more time per episode than its actual length (pausing to save think and discuss with AI) BUT the depth of knowledge I retain is incredible. This is a great example of well-designed AI-human interaction. You're still doing the work like listening reflecting deciding what resonates with you. Their AI enhances your thinking rather than replacing it. Amazing product @snipd_app @KevinBenSmith 👏"
X Link 2025-12-07T02:28Z 9889 followers, 1104 engagements

"Manus was one of the few experiences for me that showed what can be achieved with well-designed agent architecture. As @peakji mentioned in the LangChain webinar they have changed the agent design several times so their learnings in context engineering are very valuable. Apart from them companies and leaders like Anthropic Cognition Chroma and @simonw have been sharing their insights publicly and this blog does a great job of compiling them while explaining Manus's context management system. I can add @dexhorthy 's presentations on YouTube and Langchain's blog to the learning resources for"
X Link 2025-12-08T20:52Z 9889 followers, 62.1K engagements

"You should NOT use LLMs to generate synthetic human-like profiles. I just read the NeurIPS paper "LLM Generated Persona is a Promise with a Catch" and it confirms a suspicion weve held for a long time: You cannot "invent" a realistic human being using just statistics and an LLM. Yes they are more scalable and cost-effective alternative to human interviews to create digital expert personas but this paper also proves that these synthetic profiles contain systematic biases that skew simulation results away from real-world outcomes. The more creative freedom you give an LLM to generate a personas"
X Link 2025-12-09T23:08Z 9889 followers, 79.6K engagements

"I havent read the paper but the key issue is updating the context from memory. I find prompting ineffective in this so I trained a model using Thinking Machines but it's not yet ready for production. Agentic context engineering with memory focused sft models seems to be the answer"
X Link 2025-12-05T23:17Z 9844 followers, XXX engagements

"Im getting this question quite often lately: how do I create UI that doesn't look like it's AI-generated without writing code myself Its all about prompt design. Persona prompts and Skills are what make a difference in my process. I attached the post below where I described it in more detail"
X Link 2025-12-07T16:56Z 9867 followers, 2452 engagements

"We need to build more benchmarks like Vending-Bench from Andon Labs to test different verticals with real viability. Im brainstorming Agency-Bench: A sim for an AI-run writing agency focused on SaaS clients. Agents handle briefs (e.g. product descriptions blogs emails) generate/iterate content simulate A/B tests for engagement manage "client" feedback loops and profit from "retainers." Starts with $5K budget (uses for outreach and tool requests) runs X "months. Would you provide support for this"
X Link 2025-12-10T07:07Z 9844 followers, XXX engagements

"@a16z OpenAI's Seed round was $120K Weird times"
X Link 2025-12-09T01:19Z 9885 followers, 30.8K engagements

"@eliasroman @nbrichtova @OfficialLoganK @brian_lovin Yes thanks for sharing it. I've started exploring Flow great for uploading the Nano Banana & Whisk frames and Gemini prompts to create scenes"
X Link 2025-12-11T15:02Z 9875 followers, XX engagements

"Turning Churchill's speechwriting style into a Copywriter Skill works unexpectedly well. Here's a mock fundraising announcement from a startup founder. Prompt:"
X Link 2025-12-11T16:20Z 9885 followers, XXX engagements

"Churchill's writing style prompt works great with Grok as well. @elonmusk announces that Grokipedia's traffic has surpassed Wikipedia for the first time in history. "They had the quiet power to label one side fringe and the other side consensus even when the consensus was wrong and the fringe turned out to be right. But they never had you. They never had the millions who got tired of being lectured tired of being censored tired of being told what to think.""
X Link 2025-12-11T16:31Z 9885 followers, XXX engagements

"Video overviews infographics and slide decks are very good but they lack customization. Using Whisk to test various styles would work well within that interface. Also a basic version of Flow would be great to animate and build short videos from the persona campaign brief and brand details that are already in the NotebookLM project"
X Link 2025-12-11T19:33Z 9885 followers, XX engagements

"Your best people can't document their expertise because they don't know what they know until they're asked. We built an interviewer that achieves peer status so experts reveal the judgment patterns they'd only share with a colleague. I wrote a blog about how we architected the multi-agent system behind this how we extract expert thinking and build digital personas that feel like talking to a peer"
X Link 2025-12-05T01:46Z 9889 followers, 245.9K engagements

"Anthropic's "Soul" document was recently leaked and they've confirmed it's real. The document heavily focuses on safety and alignment but there's a lot to learn here about character training. I'll share my findings in three categories: Prompt Engineering Persona Embodiment and Agentic Behaviors. # Prompt Engineering Techniques The document moves beyond simple "do this don't do that" instructions. Instead it uses mental models to guide the AI's judgment. X. Hierarchical Priority Stack Instead of giving Claude a flat list of rules Anthropic establishes a numbered priority order: safety ethics"
X Link 2025-12-05T16:34Z 9887 followers, 28K engagements

"AI Agent Personas should simulate the structure of human reasoning. Ive been arguing that you cannot "invent" a digital expert agent using just prompt engineering. You have to extract the expert via deep interviewing. A new NeurIPS paper "Simulating Society Requires Simulating Thought" reinforces everything we've discussed about why thin synthetic LLM personas fail. Most AI agents operate as "behaviorists." When you prompt an LLM to "act like a senior economist" it relies on surface-level correlations from training data. It generates text that sounds expert-like but lacks any internal belief"
X Link 2025-12-11T18:58Z 9889 followers, 31.7K engagements

"@OpenAI @btibor91 Designs are amazing"
X Link 2025-12-11T23:59Z 9889 followers, 29K engagements