ChatGPT Usage Tips

Last week, I described four design patterns for AI agentic workflows that I believe will drive significant progress: Reflection, Tool use, Planning and Multi-agent collaboration. Instead of having an LLM generate its final output directly, an agentic workflow prompts the LLM multiple times, giving it opportunities to build step by step to higher-quality output. Here, I'd like to discuss Reflection. It's relatively quick to implement, and I've seen it lead to surprising performance gains. You may have had the experience of prompting ChatGPT/Claude/Gemini, receiving unsatisfactory output, delivering critical feedback to help the LLM improve its response, and then getting a better response. What if you automate the step of delivering critical feedback, so the model automatically criticizes its own output and improves its response? This is the crux of Reflection. Take the task of asking an LLM to write code. We can prompt it to generate the desired code directly to carry out some task X. Then, we can prompt it to reflect on its own output, perhaps as follows: Here’s code intended for task X: [previously generated code] Check the code carefully for correctness, style, and efficiency, and give constructive criticism for how to improve it. Sometimes this causes the LLM to spot problems and come up with constructive suggestions. Next, we can prompt the LLM with context including (i) the previously generated code and (ii) the constructive feedback, and ask it to use the feedback to rewrite the code. This can lead to a better response. Repeating the criticism/rewrite process might yield further improvements. This self-reflection process allows the LLM to spot gaps and improve its output on a variety of tasks including producing code, writing text, and answering questions. And we can go beyond self-reflection by giving the LLM tools that help evaluate its output; for example, running its code through a few unit tests to check whether it generates correct results on test cases or searching the web to double-check text output. Then it can reflect on any errors it found and come up with ideas for improvement. Further, we can implement Reflection using a multi-agent framework. I've found it convenient to create two agents, one prompted to generate good outputs and the other prompted to give constructive criticism of the first agent's output. The resulting discussion between the two agents leads to improved responses. Reflection is a relatively basic type of agentic workflow, but I've been delighted by how much it improved my applications’ results. If you’re interested in learning more about reflection, I recommend: - Self-Refine: Iterative Refinement with Self-Feedback, by Madaan et al. (2023) - Reflexion: Language Agents with Verbal Reinforcement Learning, by Shinn et al. (2023) - CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing, by Gou et al. (2024) [Original text: https://lnkd.in/g4bTuWtU ]

STOP asking ChatGPT to "make it better". Here's how to better prompt it instead: ☑ Clearly Identify the Issue Rather than a vague “make it better,” specify the exact element that needs change. For example: "Rewrite the second paragraph so it includes three concrete examples of our product’s benefits. The tone must be formal and persuasive. Remove any informal language or redundant phrases." ☑ Divide the Task into Discrete Steps Break the overall revision into a sequence of manageable tasks. For example: "Go through my instructions, step by step. – Step 1: Summarize it in one sentence. – Step 2: Identify two specific weaknesses. – Step 3: Rewrite the text to address these weaknesses, incorporating specific data or examples." ☑ Specify the Format and Level of Detail Define exactly how the final output should look. For example: "Provide the final revised text as a numbered list where each item contains 2–3 sentences. Each item must include at least one statistical fact or concrete example, and the overall response should not exceed 250 words." ☑ Request a Chain-of-Thought Explanation Ask the model to detail its reasoning process before giving the final output. For example: "Before providing the final revised text, explain your reasoning step-by-step. Identify which parts need improvement and how your changes will enhance clarity and professionalism. Then, present the final revised version." ☑ Conditional Instructions to Enforce Compliance Add if/then conditions to ensure all requirements are met. For example: "If the revised text does not include at least two concrete examples, then add a sentence with a real-world statistic. Otherwise, finalize the response as is." ☑ Consolidate All Instructions into One Prompt Integrate all the detailed instructions into a single, comprehensive prompt. For example: "First, identify the section of the text that needs improvement and explain why it is lacking. Next, summarize the current text in one sentence and list two specific weaknesses. Then, rewrite the text to address these weaknesses, ensuring the revised version includes three concrete examples, uses a formal and persuasive tone, and is structured as a numbered list with each item containing 2–3 sentences. Each list item must include at least one statistical fact or example, and the overall response must be no longer than 250 words. Before providing the final text, explain your reasoning step-by-step. If the revised text does not include at least two concrete examples, add an additional sentence with a real-world statistic." ___ Why This Works People never give enough context. And once ChatGPT answers, they never correct it enough. Think about it like an intern. Deep prompting is all about precision: give clear instructions, context & the right corrections. PS: Don't forget to use the new o3-mini model. It's crushing any other one. Yes – even DeepSeek.

I read forty LinkedIn posts before breakfast this morning. Thirty-seven sounded identical. The three that didn't shared one thing: a voice trained on the person, not the model. 📢 Generic AI drafting gives you the model's default house style. Ask any tool to "write a LinkedIn post about X" with no brand input, and you get the same six sentence shapes every other unbranded prompt produces — the dramatic one-liner, "here's the kicker," "delve into the landscape." Strip the name off any of them and you cannot tell which founder wrote it, because none of them did. 1️⃣ First, the fix isn't better prompting. It's feeding the model your actual sentences to imitate: old posts, real cadence, the words you'd never say, the ones you overuse on purpose. 2️⃣ Second, a model trained on your voice writes in your rhythm, not its default one. The output stops sounding like content and starts sounding like you had a good morning and typed fast. 3️⃣ Third, the payoff shows up in the comments, not the impressions. Generic posts collect polite likes from people skimming. Voice-true posts get replies that argue back, because the reader believes a specific person wrote the sentence. A model can draft your post. Only your voice makes someone stop scrolling. Every client content pipeline I run now starts with a brand voice file, before a single post gets drafted. What's the last AI-written post you read that actually sounded like the person, not like every other AI-written post?

I've been building and deploying RAG systems for 2+ years. And it's taught me optimizing them requires focusing on 3 core stages: 1. Pre-Retrieval 2. Retrieval 3. Post-Retrieval Let me explain - Most people focus on the generation side of things. But optimizing retrieval is what really makes the difference. Here's how to do it: 𝟭/ 𝗣𝗿𝗲-𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 This is where we optimize the data before the retrieval process even begins. The goal? Structure your data for efficient indexing and ensure the query is as precise as possible before it's embedded and sent to your vector DB. Here’s how: - 𝗦𝗹𝗶𝗱𝗶𝗻𝗴 𝘄𝗶𝗻𝗱𝗼𝘄: 𝘐𝘯𝘵𝘳𝘰𝘥𝘶𝘤𝘦 𝘤𝘩𝘶𝘯𝘬 𝘰𝘷𝘦𝘳𝘭𝘢𝘱 𝘵𝘰 𝘳𝘦𝘵𝘢𝘪𝘯 𝘤𝘰𝘯𝘵𝘦𝘹𝘵 𝘢𝘯𝘥 𝘪𝘮𝘱𝘳𝘰𝘷𝘦 𝘳𝘦𝘵𝘳𝘪𝘦𝘷𝘢𝘭 𝘢𝘤𝘤𝘶𝘳𝘢𝘤𝘺. - 𝗘𝗻𝗵𝗮𝗻𝗰𝗶𝗻𝗴 𝗱𝗮𝘁𝗮 𝗴𝗿𝗮𝗻𝘂𝗹𝗮𝗿𝗶𝘁𝘆: 𝘊𝘭𝘦𝘢𝘯, 𝘷𝘦𝘳𝘪𝘧𝘺, 𝘢𝘯𝘥 𝘶𝘱𝘥𝘢𝘵𝘦 𝘥𝘢𝘵𝘢 𝘧𝘰𝘳 𝘴𝘩𝘢𝘳𝘱𝘦𝘳 𝘳𝘦𝘵𝘳𝘪𝘦𝘷𝘢𝘭. - 𝗠𝗲𝘁𝗮𝗱𝗮𝘁𝗮: 𝘜𝘴𝘦 𝘵𝘢𝘨𝘴 (𝘭𝘪𝘬𝘦 𝘥𝘢𝘵𝘦𝘴 𝘰𝘳 𝘦𝘹𝘵𝘦𝘳𝘯𝘢𝘭 𝘐𝘋𝘴) 𝘵𝘰 𝘪𝘮𝘱𝘳𝘰𝘷𝘦 𝘧𝘪𝘭𝘵𝘦𝘳𝘪𝘯𝘨. - 𝗦𝗺𝗮𝗹𝗹-𝘁𝗼-𝗯𝗶𝗴 (or parent) 𝗶𝗻𝗱𝗲𝘅𝗶𝗻𝗴: 𝘜𝘴𝘦 𝘴𝘮𝘢𝘭𝘭𝘦𝘳 𝘤𝘩𝘶𝘯𝘬𝘴 𝘧𝘰𝘳 𝘦𝘮𝘣𝘦𝘥𝘥𝘪𝘯𝘨 𝘢𝘯𝘥 𝘭𝘢𝘳𝘨𝘦𝘳 𝘤𝘰𝘯𝘵𝘦𝘹𝘵𝘴 𝘧𝘰𝘳 𝘵𝘩𝘦 𝘧𝘪𝘯𝘢𝘭 𝘢𝘯𝘴𝘸𝘦𝘳. - 𝗤𝘂𝗲𝗿𝘆 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻: 𝘛𝘦𝘤𝘩𝘯𝘪𝘲𝘶𝘦𝘴 𝘭𝘪𝘬𝘦 𝘲𝘶𝘦𝘳𝘺 𝘳𝘰𝘶𝘵𝘪𝘯𝘨, 𝘲𝘶𝘦𝘳𝘺 𝘳𝘦𝘸𝘳𝘪𝘵𝘪𝘯𝘨, 𝘢𝘯𝘥 𝘏𝘺𝘋𝘌 𝘤𝘢𝘯 𝘳𝘦𝘧𝘪𝘯𝘦 𝘵𝘩𝘦 𝘳𝘦𝘴𝘶𝘭𝘵𝘴. 𝟮/ 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 The magic happens here. Your goal is to improve the embedding models and leverage DB filters to retrieve the most relevant data based on semantic similarity. - Fine-tune your embedding models or use instructor models like instructor-xl for domain-specific terms. - Use hybrid search to blend vector and keyword search for more precise results. - Use GraphDBs or multi-hop techniques to capture relationships within your data. 𝟯. 𝗣𝗼𝘀𝘁-𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 At this stage, your task is to filter out noise and compress the final context before sending it to the LLM. - Use prompt compression techniques. - Filter out irrelevant chunks to avoid adding noise to the augmented prompt (e.g., using reranking) 𝗥𝗲𝗺𝗲𝗺𝗯𝗲𝗿: RAG optimization is an iterative process. Experiment with various techniques, measure their effectiveness, compare them and refine them. Ready to step up your RAG game? Check out the link in the comments.

One of the biggest challenges I see with scaling LLM agents isn’t the model itself. It’s context. Agents break down not because they “can’t think” but because they lose track of what’s happened, what’s been decided, and why. Here’s the pattern I notice: 👉 For short tasks, things work fine. The agent remembers the conversation so far, does its subtasks, and pulls everything together reliably. 👉 But the moment the task gets longer, the context window fills up, and the agent starts forgetting key decisions. That’s when results become inconsistent, and trust breaks down. That’s where Context Engineering comes in. 🔑 Principle 1: Share Full Context, Not Just Results Reliability starts with transparency. If an agent only shares the final outputs of subtasks, the decision-making trail is lost. That makes it impossible to debug or reproduce. You need the full trace, not just the answer. 🔑 Principle 2: Every Action Is an Implicit Decision Every step in a workflow isn’t just “doing the work”, it’s making a decision. And if those decisions conflict because context was lost along the way, you end up with unreliable results. ✨ The Solution to this is "Engineer Smarter Context" It’s not about dumping more history into the next step. It’s about carrying forward the right pieces of context: → Summarize the messy details into something digestible. → Keep the key decisions and turning points visible. → Drop the noise that doesn’t matter. When you do this well, agents can finally handle longer, more complex workflows without falling apart. Reliability doesn’t come from bigger context windows. It comes from smarter context windows. 〰️〰️〰️ Follow me (Aishwarya Srinivasan) for more AI insight and subscribe to my Substack to find more in-depth blogs and weekly updates in AI: https://lnkd.in/dpBNr6Jg

This Is The BEST ChatGPT Prompt Formula (For Job Seekers): 70%+ of job seekers are using AI in their job search. But most of them use the same basic prompts, which lead to generic outputs that don’t get results. I’ve spent 100+ hours this year engineering prompts, then tested them with the hundreds of clients in our job search coaching program. Here is the formula we’ve found to be most effective: Part 1: Context Context is an overview of the scenario including relevant information and data so ChatGPT can understand and assess. For example: ❌ Bad: Please rewrite my resume bullets. ✅ Good: I am revising my resume with the goal of applying for Account Executive roles at F500 tech companies. I’m attaching the job description of an example role I’d be interested in. I’m also attaching a copy of my resume. I want to revise my bullets to be optimized for these types of roles. Part 2: Role This is where you define the role that ChatGPT should assume when completing this task for you. Build your unicorn. This “persona” can have a PhD in marketing AND 20+ years as a recruiter who is super data-driven and numbers focused. ❌ Bad: Skipping the role entirely (most do this) ✅ Good: Assume the role of a data-driven resume writer with 20+ years of experience as a hiring manager and recruiter for F500 technology companies like Salesforce, Amazon, and Microsoft. Part 3: Actions This is where you outline the specific actions that ChatGPT should take to complete your task. If you’re not sure, in a separate chat, ask ChatGPT: “Please give me detailed steps for [Task] to achieve [Outcome].” Then review and tweak those steps. ❌ Bad: Please rewrite this resume bullet. ✅ Good: Scan the job description to identify relevant keywords and qualifications, as well as goals and challenges for this role. Review my resume to get an understanding of my experience, qualifications, skills, and achievements. Revise each bullet on my resume so it includes keywords from my target roles, includes measurable outcomes, uses compelling language, and is between 12–20 words. Share the revised bullets along with a summary of the changes you made for each role in the Experience section of my resume. Part 4: Examples Share an example of what a “good” output should look like for this task. Not sure? Run the prompt without this part, but ask it to share 10 variations. Review the 10 outputs, then mix and match to create your “perfect output.” Now revise your main prompt with it. ❌ Bad: Here’s a bullet from my resume: [Bullet] ✅ Good: Here is an example of what a good output should look like: Austin’s Company Account Executive Surpassed quarterly quotas by 115% through strategic client relationships and closing multi-year SaaS contracts. Owned full sales cycle, generating $1.3M in annual recurring revenue across SMB, mid-market, and enterprise accounts. Give it a shot for your Resume, Cover Letters, LinkedIn, Networking, and anything else in your search!

"Using ChatGPT in your job search is dangerous!" Wrong. It's called being smart. Top candidates are using ChatGPT to make their search faster and easier. Let's be clear: If you are job hunting, your immediate goal is the interview. You won't land an interview if: ❌ Your resume isn't tailored to each job posting ❌ Your resume isn't in an ATS-friendly format ❌ Your cover letter (yes, you need one!) is generic ❌ You haven't tied your skills to the job requirements 6 ways to use ChatGPT in your job search, along with prompts: 1️⃣ Resume Tailoring Prompt: "Review my resume attached, and suggest edits to tailor it to the job posting attached. Suggest bullets beginning with actions verbs. Highlight any opportunities to support my skills with data or quantifiable achievements." 2️⃣ LinkedIn Profile Optimization Prompt: "I am looking for a role as [job title]. Based on my resume and the job posting (attached), write a LinkedIn About section that shows my abilities to fill the role of [job title]. Do not include anything that is not in my resume. Do not embellish my background or assume I have additional qualifications. Then summarize this and suggest 3 compelling LinkedIn Headline alternatives." 3️⃣ Cover Letters Prompt: "Here is a job posting (attached), and my resume (attached). Please draft a compelling cover letter that highlights my qualifications for this job. Use strong hook in the introductory paragraph, followed by bullet points starting with action verbs describing my skills. Be sure to tie my background to the needs described in the posting." 4️⃣ Company Research Prompt: "I am interviewing for a job as [title] at [company]. Describe the current company business model, mission, and biggest business challenges. Identify how this role fits into the company structure. Describe the company culture and [add any of your personal needs/prioritiies here, like maternity leave, DEI, etc.] 5️⃣ Interview Prep Prompt: "I am interviewing for the job described in the attached posting at [name of company]. Please identify at least 15 questions that I can expect to be asked in the interview for this job, and provide sample answers a STAR format. Include any company-specific questions I can expect." 6️⃣ Salary Talks Prompt: "I am looking a roles as a [job title] in the [industry]. I have X years of experience. What is a reasonable salary range for someone with my experience in this role? Write a script for me to use to negotiation the top of that range that focuses on my skills and what my market value is. Include any points I should be prepared to negotiate or to compromise on based on the resume attached." 🚨CAUTION! 🚨  ➙ Always put responses in your own voice!  ➙ AI is your assistant, not your replacement. Proofread! What's your favorite AI strategy right now? ♻️ Share to help others update their job search 🔔 Follow Sarah Baker Andrus for more evidence-based career strategies 📌 Struggling with your search? DM me to chat.

Something has shifted in how reliable ChatGPT is for professional work. It is not perfect. But the gap between AI output and usable output has narrowed considerably. Here is an honest look at what GPT-5.5 is actually good for, and where caution still applies. The reliability question Hallucinations have not been eliminated. But the improvement is real and worth understanding. OpenAI attributes the reduction not to a smarter base model, but to behavioural changes during work: better tool use, checking its own work, grounded search, and post-training penalties for overconfident wrong answers. In practical terms, the model is more likely to flag uncertainty rather than fill gaps with plausible-sounding fiction. The caveat: citation accuracy remains the weakest area — models still invent references, paper titles, and author names at a meaningful rate. If your work involves sourcing or referencing, verify independently. What it is genuinely strong at Research and synthesis. GPT-5.5 handles large volumes of information well — documents, reports, transcripts — and organises them into structured, usable summaries. The reasoning across long material has improved significantly. Analysis and structured thinking. It works through complex, multi-part problems with greater consistency than previous versions. Strategy questions, scenario planning, and professional problem-solving across law, finance, and business are all areas where it now adds real value. Document and report drafting. The writing quality is strong. For professionals who need a capable first draft — proposals, briefs, analysis documents — it reduces time spent considerably. What it still is not A replacement for judgement. The output quality depends entirely on the quality of the thinking behind the prompt. Vague inputs produce vague outputs. A reliable source of citations or references. Always verify. A substitute for domain expertise on high-stakes decisions. Use it to think through a problem, not to make the final call. The shift worth paying attention to The question most professionals were asking two years ago was: can I trust this? The more useful question now is: am I using this for the right tasks? GPT-5.5 is a strong tool for knowledge work. The professionals getting the most from it are not the ones using it most — they are the ones using it most deliberately. #AIStrategy #ChatGPT #AITools #BusinessProductivity #MarketingAI

As an ML Engineer, I deal a lot with Prompt Engineering to get the best result from LLMs. With that exp. I have created the roadmap about how to learn Promot Engineering and write the best prompt: 1/ Understand How LLMs Work - LLMs predict the next token, not “truth” - They’re trained on massive text corpora - Everything depends on the context you give them - If your prompt lacks structure → your output lacks accuracy. 2/ Start with Prompt Basics - Great prompts are clear, structured, and instructive - Use explicit instructions: “Summarize this in 3 bullet points” - Add role/context: “You are a data scientist…” - Be specific with constraints: “Limit answer to 100 words” - Avoid vague prompts like: “Tell me about LLMs” 3/ Practice Prompting Styles - Explore different prompting techniques - Zero-shot: Just ask the question - Few-shot: Add examples to guide the model - Chain-of-thought: Ask the model to “think step by step” - Self-refinement: “What could be improved in the above?” - These patterns reduce hallucinations and improve quality. 4/ Explore Real-World Use Cases - Summarizing long documents - Extracting insights from PDFs or tables - Building a chatbot with memory - Writing job descriptions, SQL queries, or ML code - Use tools like LangChain, LlamaIndex, or PromptLayer for structured experiments. 5/ Learn from Experts - OpenAI Cookbook - Prompt Engineering Guide (awesome repository on GitHub) - Papers like "Self-Instruct", "Chain-of-Thought Prompting", "ReAct" - Courses: Deeplearning . ai’s "ChatGPT Prompt Engineering" (by OpenAI) 6/ Document Your Best Prompts - Test iteratively - A/B test prompts to find the most effective version - Note what works (or fails) - Build your own prompt library! 7/ Automate & Deploy - Use APIs (OpenAI, Claude, Gemini) in Python - Build apps using Streamlit + LLMs - Store embeddings using FAISS or ChromaDB - Build Retrieval-Augmented Generation (RAG) pipelines One of my bonus tip: Use AI to write more refined prompt. Sounds weird? - First, document what you require - ask AI to generate an AI friendly prompt for best result - and see the results - 10x better than your own prompt! In the LLM era, your prompt is your superpower. Repost this if you find it useful. #ai #ml #prompt #llm

Stop filling your agent's context window just because you can. A few months ago, I worked on a browser agent which used Playwright MCP to navigate career pages. Upon integrating the MCP server, I noticed an interesting problem. The agent sometimes picked the wrong tools for navigation. When I dug further, it started to make sense. Playwright MCP offers 26 tools. Most of which aren't relevant to my workflow.  I needed my agent to fill forms, click links, etc. I didn't need a browser_network_request or browser_file_upload tool. In fact, I only needed 8 tools but my browser agent didn't know that. It took the presence of all 26 tools as a license to potentially use any of them. The fix was simple. I filtered down the tools to the few I needed, and I got better performance immediately. At the time, I didn't have the words to describe this problem until I read an article by Drew Breunig. Drew argues that even though modern LLMs have large context windows, we should be intentional about what goes in. In my case, my agent had fallen prey to what he calls '𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗖𝗼𝗻𝗳𝘂𝘀𝗶𝗼𝗻' - when unnecessary context is used by the agent, degrading its decision-making over time. Aside from '𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗖𝗼𝗻𝗳𝘂𝘀𝗶𝗼𝗻', Drew identified three other failure modes:     1. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗖𝗹𝗮𝘀𝗵: This happens when data from different sources returns conflicting results. The agent then makes wrong inferences based on this. A common example is a coding agent which pulls information from two sources: official docs and say an outdated blog post. The agent can potentially use the outdated post or even synthesise a new wrong idea of how the library should work based on both sources.     2. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗣𝗼𝗶𝘀𝗼𝗻𝗶𝗻𝗴: This happens when an error, outdated data or even hallucination from the LLM makes it into the context. The LLM goes through this info and potentially uses it in generating answers, thus perpetuating the error. Imagine running a multi-step agent where the model hallucinates, say, a product name or detail. That summary gets passed on as context to the next step. From that point, every further output is built on the wrong fact. 3. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗗𝗶𝘀𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻: the agent over-relies on past behaviour, responses and interactions rather than reasoning afresh based on what the user needs. All these failure modes point to a simple idea - give the LLM what it needs to make the right decisions and nothing more. Context is not a dumping group and what goes in shapes what comes out of your agents. Of course, this simple idea involves a lot more design and engineering upfront. There's even an entire field (Context Engineering) built on top and I'll be sharing more of my learnings so stay tuned! :) Now I'm keen to know, which of these failure modes have you encountered and how did you fix them? Share in the comments!