About
Most…
Articles by Marc
Activity
4K followers
-
Marc G. shared thisHow about you git repoMarc G. shared thisHow to build AI Agents from scratch even if you’ve never done it before. I've spent the last few months building agents with Claude Code. And the same 7 steps that used to take me weeks now take an afternoon. (Save this before you build your first one 👇) 1️⃣ Define your use case ↳ One problem. One success metric. Clear boundaries. ↳ Bad scoping = broken agent. Always. 2️⃣ Pick your scaffolding ↳ LangChain, CrewAI, Vercel AI SDK, LlamaIndex... or Claude Code. ↳ Pick on complexity, not hype. 3️⃣ Connect your LLM ↳ Claude, GPT, Gemini. Configure params. Set keys. Test responses. ↳ Your agent is only as smart as its brain. 4️⃣ Add tools ↳ Web search. Email. DB queries. APIs. ↳ No tools = no action = just a chatbot. 5️⃣ Add memory ↳ Short-term for the conversation. Vector DB for the long game. ↳ Without it, your agent forgets everything after one turn. 6️⃣ Add a loop ↳ ReAct. Plan-and-Execute. Tree of Thoughts. ↳ This is what controls how your agent thinks and loops. 7️⃣ Add guardrails ↳ Rate limits. Input/output filtering. Human-in-the-loop. AI Evals. ↳ Skip these and your agent does something you can't explain. Then test, deploy, monitor, improve. That's the full loop. Most beginners skip steps 5, 6 and 7. Then wonder why their agent hallucinates, loops forever, or burns $300 in API calls in one weekend. That's exactly why I built this LinkedIn Learning course: Vibecoding Your AI Agent with Claude Code. It's available for FREE here: https://lnkd.in/g_4CrfE7 Which step do you think most builders skip first? Drop it in the comments below. ♻️ Repost to help builders in your network ship their first agent. And follow Basia Kubicka for more on building AI with Claude Code. TY GenAI Works for great infographic.
-
Marc G. reposted thisMarc G. reposted this✨ AI at EA Spotlight Series: Intern Edition Ft. Allyson Greenberg and Brendan Fielman At EisnerAmper, innovation is a core part of our culture, but it must always be grounded in responsible AI use and the protection of our clients’ data. This summer, interns Ally and Brendan are partnering with our AI Governance Lead, Dave O'Brien, to support firmwide governance campaigns that help our people understand how to experiment, build, and innovate with AI safely and responsibly. Their work is a great example of how colleagues at every role can play a meaningful role in shaping the future of AI at our firm. EisnerAmper David Frigeri Dave O'Brien Brian Skelly Michelle Bradley, M.Ed. Jalen Gonel
-
Marc G. shared thisThree posts. One engine. First, Steve Jobs in 1985 — you can read Aristotle, but you can't ask him a question. Then I built the tool he described. With Claude, in about an hour. Jobs qualified for his own roster, so I asked him what he thought. "The costume is easy. The engine is hard. Always ship the engine." Then Turing crashed the room and ran his test. Nobody could tell. Now the whole engine is documented — the methodology, the guardrails, the clever part nobody noticed: every layer of this project demonstrates the method it documents. The tool channels the man who imagined it. The blog was drafted inside the archive it describes. Full story here → https://lnkd.in/ewa7qGcx And the roster just got its first difficult addition: Nietzsche. His test for any life was eternal recurrence — would you will it again, unchanged, forever? The build took an hour. Yes. Forever. Obviously. Read it. Fork it. Add your Aristotle. https://lnkd.in/eBEmT52C #AI #SteveJobs #TuringTest #OpenSource #Braintrailz
-
Marc G. shared thisNobody could tell. I built a Steve Jobs AI persona with Claude last week — in about an hour, on a Mac mini. Then I let the rest of the OGs into the room: Fred Brooks, still muttering about man-months. The C crowd, arguing over semicolons. And then, uninvited but right on time — Alan Turing. He'd asked "can machines think?" thirty-five years before Jobs ever dreamed of putting a mind in a box. Jobs' persona offered him a chair. Turing didn't sit. He pulled out a notepad and asked the room to guess, blind, which answers were his and which were the machine's. Nobody could tell. Turing finally sat down. "Indistinguishable," he said. "That's the whole result — nothing more, nothing less." The costume is easy. The engine is hard. Who would you let into the room next? #AI #TuringTest #SteveJobs #TechHistory #Anthropic
-
Marc G. shared thisstart with this https://lnkd.in/exms_kce then look at tis https://lnkd.in/eDbn-mJM Ever wondered what happens when you mix Steve Jobs’ vision with the "Mythical Man-Month" and the creators of C? https://lnkd.in/eBEmT52C I just dropped a post connecting the OGs of tech to the AI era. You’ve gotta see this
-
Marc G. shared thisFollow-up to this morning’s post. Most AI personas put on a toga and call themselves Aristotle. I wanted the thinker, not the costume. In 1985, Steve Jobs imagined capturing a great mind’s worldview in a computer so we could ask it questions. Today, I built it—with Claude, in about an hour. https://lnkd.in/egEMJ33K Then came the twist: Steve Jobs qualifies for the tool Steve Jobs imagined. So I asked what he thought about it. “The costume is easy. The engine is hard. Always ship the engine.” Free. Open source. Running on a Mac mini—obviously. https://lnkd.in/eBEmT52C Who would you channel first? Like it. Repost it. Follow me. Let’s give LinkedIn a real AI hot take. #AI #SteveJobs #OpenSource #Braintrailz
-
Marc G. shared thisI had the opportunity to meet Steve at a Novell Brainshare conference during his 'Next' years. I often think about what foresight and pure genius means. I Just watched this twice. The first time with a giant smile and the second time with a small tear. https://lnkd.in/eYw9696NMarc G. shared thisSteve Jobs said that when a new medium appears, we usually understand it through the old one. Brilliant Share By: Tom Goodwin Original Post Below 👇 👇 👇 Steve Jobs said that when a new medium appears, we usually understand it through the old one. In 1985, he predicted generative AI. He was right about the technology. He was also right about the mistake we'd make with it. Computers were treated like typewriters and the internet was treated like the newspaper. Each time, we took something new and mapped it onto something familiar. When you see something new through the lens of something old, you only ever use it to do what you were already doing, just faster. You never ask what becomes possible that wasn't possible before. The same thing is happening with AI. Most people use AI only as a productivity tool: faster writing, better summaries, quicker research. That's real value, but it's the surface layer. A smaller group use it as an operating layer: workflows, decision frameworks, sales intelligence. But there's a consumer side that isn't getting enough attention yet. A growing segment of shoppers are already using it to buy things. Not searching Google and clicking through to a product page. Asking ChatGPT for a product and getting a recommendation with a link. That shift is small right now. But I've watched behaviour change enough times to know what early looks like. I did a post the other day about how my kids don't use Google, but mostly the voice mode in ChatGPT. What they do today is what every shopper does in five years. So here's what I think will happen: The next three years move AI shopping from something a minority of consumers experiment with to the default starting point for high-consideration purchases. Trust won't just convert shoppers, but will decide which brands AI recommends. Steve Jobs imagined a tool that could capture the greatest minds in history and make them interactive. We built it. In terms of how we use it, we have only scratched the surface. ♻️ Repost for a leader building for what's coming. 💡 Follow Tom Goodwin for more on building the path to agentic commerce. ________________________ DM us to be featured on our next post Master the language of AI. One prompt at a time. How to Talk to AI
-
Marc G. shared thisIn God We Trust, Everybody Else Bring Data... After reviewing the recent jobs report I wondered how that aligned with my own data on the subject. Here is what boiled up to the top. 1. The "Low-Hire, Low-Fire" Gridlock & High 2. High Fluctuation and Invalidation in Tech Giants 3. The Structural Shift: AI & Infrastructure Dominance 4. Cross-Industry Modernization I checked if 25 director and VP AI postings were actually real. Not screenshotted the title. Opened every one. 18 were dead. Filled, 404, closed. Seventy two percent. Visa, Amex, CrowdStrike, SAP, and Lenovo all had listings that looked alive and weren't. The 7 that were real had nothing to do with the companies everyone's obsessing over. No OpenAI. No hot startup with a demo and no product. Florida Blue. Mazda. Foot Locker. SavATree. Atos. Valiantys, twice. Healthcare, automotive, retail, field services, consulting. All companies sitting on legacy ERP and decades of data debt with a CIO who just got budget to fix it. If your pitch is "I build agents," you're fighting bootcamp grads for scraps at companies burning venture money. If it's "I modernize enterprise architecture and know how to wire AI into it without breaking the business," you're one of maybe a dozen people they can actually hire. Everyone's chasing the sexy job. The boring job is the one that's open. Tell me I'm wrong. #EnterpriseArchitecture #AI #TechLeadership
-
Marc G. shared thisBefore twitter and X the word grok was defined as to understand something so completely and intuitively that the observer becomes a part of the observed. I built a workflow to find new opportunities and have been running it in claude coworker. Heres what I learned Most AI automation is just a very confident intern with a browser. It finds 20 things. 12 are wrong. 4 are stale. 4 might be useful. Then it makes a beautiful report. That is not intelligence. That is a PowerPoint-shaped hallucination. The hard part is not retrieval. The hard part is knowing what to ignore. AI does not need more prompts. It needs more fact-checking.
-
Marc G. reacted on thisMarc G. reacted on thisGetting our groove on at the Square booth. Looks like this robot has some serious moves. There's no arguing with the fact that Chad and I do a lot of events. It's the silly moments like this that fill the spaces between the serious conversations that recharge our batteries and sharpen our axes. #rspa #retailnow #inctankgtm #chatgtmpodcast
-
Marc G. reacted on thisMarc G. reacted on this✨ AI at EA Spotlight Series: Intern Edition Ft. Allyson Greenberg and Brendan Fielman At EisnerAmper, innovation is a core part of our culture, but it must always be grounded in responsible AI use and the protection of our clients’ data. This summer, interns Ally and Brendan are partnering with our AI Governance Lead, Dave O'Brien, to support firmwide governance campaigns that help our people understand how to experiment, build, and innovate with AI safely and responsibly. Their work is a great example of how colleagues at every role can play a meaningful role in shaping the future of AI at our firm. EisnerAmper David Frigeri Dave O'Brien Brian Skelly Michelle Bradley, M.Ed. Jalen Gonel
-
Marc G. liked thisMarc G. liked this
Experience & Education
-
Beyond Limited Thinking
View Marc’s full experience
See their title, tenure and more.
or
Licenses & Certifications
Patents
-
Framework for adapter used in enterprise application integration
Issued US 7421709
A method of making adapters for enabling incompatible computer systems to work in cooperation by providing a framework of software rules and resources from which to configure adapters for mediating between incompatible applications. The adapters are each provided with at least one interface for receiving requests and another interface for sending information. The framework resources and rules provide objects which can be reused for making one-way adapters, bi-directional adapters, round-trip…
A method of making adapters for enabling incompatible computer systems to work in cooperation by providing a framework of software rules and resources from which to configure adapters for mediating between incompatible applications. The adapters are each provided with at least one interface for receiving requests and another interface for sending information. The framework resources and rules provide objects which can be reused for making one-way adapters, bi-directional adapters, round-trip adapters, and compound adapters.
Other inventors -
-
Framework for adapter used in enterprise application integration.
20050149941
Projects
-
Gridlinks - A Open Source Chrome Extension
A chrome extension framework build in reactjs that gets it data using genai and presents a grid full of links. Open source and ready for reuse.
-
iWay Trading Partner Manager
-
In order to save money by not hiring consultants, I was tasked with a front-end rewrite for a 8 year old product that enabled integration among Trading Partners. The original user interface was built using Ext JS Library 2.2.1, a library with a well thought out and comprehensive MVC frameworks around. It was typically used for developing data-centric intranet applications, where you would have used a VB or maybe a WinForms application
The rules were simple, create a drop-in replacement…In order to save money by not hiring consultants, I was tasked with a front-end rewrite for a 8 year old product that enabled integration among Trading Partners. The original user interface was built using Ext JS Library 2.2.1, a library with a well thought out and comprehensive MVC frameworks around. It was typically used for developing data-centric intranet applications, where you would have used a VB or maybe a WinForms application
The rules were simple, create a drop-in replacement with little (or no) backend changed that could run well in todays major browsers.
Technologies chosen:
* Microsoft Typescript http://www.typescriptlang.org/
* UIKit js http://getuikit.com/
* Node, Grunt, Bower,Tsd
* and Bitbucket (formerly known as Stash) for on premise git repoOther creators -
-
Enterprise Monitoring and Management.
-
iWay Sentinel. Designed the overall architecture, Bootstrap ScrumMaster, Designed and Developed the Front-End (Javascript, bootstrap 3, JQuery)
Other creators -
-
Omni Integrators Workbench On Docker
-
Information Builders, Omni Applications are domain specific master data applications that combine a enterprise mastered index with one or more pre-packaged models to provide full identity management, to achieve a 360-degree view of a key entities, with a single golden record for each player.
The Omni Integrators Workbench on Docker (omnidock) is a defined process that can be used to run standalone validation of an integrators work product.Other creators -
-
Data Integrity
-
Egret is a power tool for working with messy data. Use it to improve data consistency, link it to data registries. Based on Open Refine
Other creators -
-
smashdata
-
simple encryption for data at rest
Languages
-
English
-
Recommendations received
8 people have recommended Marc
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content