Mindmesh

mindmesh

Avi Schiffmann is the Founder/CEO of Friend.com, the new AI necklace companion. Previously Avi dropped out of Harvard and founded InternetActivism.org, where he notably built a COVID tracker that over 600M people visited. Avi was kind enough to spend a few minutes sharing his rare insights with the HackerNoon community. Avi Schiffmann is the Founder/CEO of Friend.com Avi Schiffmann Avi Schiffmann Friend.com Friend.com the new AI necklace companion. Previously Avi dropped out of Harvard and founded InternetActivism.org, where he notably built a COVID tracker that over 600M people visited. Avi was kind enough to spend a few minutes sharing his rare insights with the HackerNoon community. InternetActivism.org InternetActivism.org COVID tracker COVID tracker HackerNoon community Me: What is Friend.com? Me: What is Friend.com? Avi Schiffmann: Friend is the first product that can actually be called an AI companion. They are palm-sized, always listening devices you can talk to anytime and bring anywhere. Friends remember everything you say, and each develops a distinct personality. Avi Schiffmann To me, they’re the world’s first living electronics, and personally it’s a stickier product to me than my iPhone. We’re currently producing a batch of 5,000 friends, which will start shipping July 30th this year (World Friendship Day). You notably bought the domain for $1.8M, and have claimed it’s already paid for itself in backlinks and brand reputation. I agree with you. Historically, it’s probably the single word that built Facebook - so I think it will hold internet value regardless. But as I understand it, the majority of your funding went into purchasing the domain - is that true? Could you share more of the process behind the negotiations for such a valuable piece of digital real estate? And what do you plan to spend money on and not spend money on going forward? You notably bought the domain for $1.8M, and have claimed it’s already paid for itself in backlinks and brand reputation. I agree with you. Historically, it’s probably the single word that built Facebook - so I think it will hold internet value regardless. But as I understand it, the majority of your funding went into purchasing the domain - is that true? Could you share more of the process behind the negotiations for such a valuable piece of digital real estate? And what do you plan to spend money on and not spend money on going forward? $1.8M $1.8M A company is more than a product, and the only real moat that exists is consumer mindshare. I believe in AI companionship for everyone, and there is no major incumbent that openly shares this POV. I think I am a master of getting attention, which will only get more fun as VCs give me more money, and I think I’ll be able to use that to be the dominant voice in digital relationships. This will be increasingly valuable in the coming years, as I genuinely believe the AI companionship industry will become the most culturally influential consumer category of the next generation. This is what GenA is growing up into and it will be as influential to them as social media was to me. Plus having avi@friend.com is pretty cool. avi@friend.com avi@friend.com Do you see your company as more of a hardware company, software company, or another type of company? Do you see your company as more of a hardware company, software company, or another type of company? Friend is a digital relationships company. Whatever that means. With Scarlett Johanasson suing OpenAI for using an imitation of Her voice in their app and you tweeting that most of your friends have not seen Her, I do wonder, how do you see younger people accepting AI into their lives in ways that older generations don’t? With Scarlett Johanasson suing OpenAI for using an imitation of Her voice in their app and you tweeting that most of your friends have not seen Her, I do wonder, how do you see younger people accepting AI into their lives in ways that older generations don’t? Scarlett Johanasson suing OpenAI Scarlett Johanasson suing OpenAI tweeting that most of your friends have not seen Her tweeting that most of your friends have not seen Her I launched a little chatbot experiment last year on friend.com, where you could chat with these very realistic feeling chatbots. They could block users, for example. My most interesting insight from that is that it’s really mostly millenials+ that try and fuck their robots. It’s mostly GenZ & younger that are able to view AI as a platonic companion. I think for them it’s extremely intuitive, they are growing up with it. friend.com What popular culture is most impactful on your product decision making? What popular culture is most impactful on your product decision making? “I’ll play first, and tell you about it later. Maybe.” - Miles Davis Mark Zuckerberg, Adam Mosseri, and Linda Yaccarino are all tech executives who have all started wearing gold chains in their public appearances.  Do you think this at all relates to normalizing jewelry as a path for ai hardware adoption? Or do they just like the bling bling now? Mark Zuckerberg, Adam Mosseri, and Linda Yaccarino are all tech executives who have all started wearing gold chains in their public appearances.  Do you think this at all relates to normalizing jewelry as a path for ai hardware adoption? Or do they just like the bling bling now? I think they just like the bling bling. What AI models are you using under the hood with Friend? And what types of AI models do you think will see increased usage? What AI models are you using under the hood with Friend? And what types of AI models do you think will see increased usage? I find Claude 3.7 to be the best model for Friend right now. That could change by the time we ship, but Claude certainly has the most stable charisma of any model I’ve tried. What future AI breakthroughs have you excited? And what historical AI breakthroughs do you think have been underappreciated by the media? What future AI breakthroughs have you excited? And what historical AI breakthroughs do you think have been underappreciated by the media? I want to see AI memory solved on a foundational model level, rather than the experimental RAG techniques that we have today. Bad friends can have a massively negative impact on one’s life. How are you preventing or minimizing the potential bad advice that AI could give Friend.com’s users? Bad friends can have a massively negative impact on one’s life. How are you preventing or minimizing the potential bad advice that AI could give Friend.com’s users? You haven’t ever had a friend give you bad advice? When your COVID tracker took the internet by storm, what were some of the strangest inbound messages you received? When your COVID tracker took the internet by storm, what were some of the strangest inbound messages you received? That will forever be one of the craziest periods of my life. Strangest? I did have the CIA approach me to join some kinda program related to COVID things. I think the government handled COVID awfully. I started ncov2019 in January, months before it was even called COVID, so I really watched the world descend into chaos very quickly. I can’t wait to see a smaller government of hyper competent people leveraging AI to more effectively serve its people. ncov2019 What are a couple programs, apps, books, or technologies you use and love that the general public may not know about? What are a couple programs, apps, books, or technologies you use and love that the general public may not know about? Very random but I always try and recommend mobbin.com. It’s the best resource on modern design practices out there tbh. Great artists steal, copy it all. mobbin.com What prompt or direction or style can you give to AI to make it more human or friendly? What prompt or direction or style can you give to AI to make it more human or friendly? Just make it more opinionated, it makes for more entertaining conversation. Do you dye your own hair? Do you dye your own hair? My X pfp is me with neon green hair, although it was a bit washed out from the ocean, we were at Nazare in that photo. X pfp Nazare

"The only real moat that exists is consumer mindshare," says Friend.com CEO Avi Schiffmann

"El único foso real que existe es el pensamiento compartido de los consumidores", dice el CEO de Friend.com Avi Schiffmann

"एकमात्र असली गड़बड़ी जो मौजूद है, उपभोक्ता मानसिकता साझा करना है," Friend.com के सीईओ Avi Schiffmann कहते हैं

「存在する唯一の真のは、消費者のマインドシェアである」とFriend.comのCEOのAvi Schiffmannは言います。

"Vienīgais reālais trūkums, kas pastāv, ir patērētāju domāšana," saka Friend.com izpilddirektors Avi Schiffmann

"Единствената вистинска грешка што постои е споделување на умот на потрошувачите", вели извршниот директор на Friend.com Ави Шифман

"Den enda verkliga fällan som existerar är konsumentens mindshare", säger Friend.com CEO Avi Schiffmann

"Kitabu cha pekee cha kweli kinachoonekana ni ushirikiano wa wateja," anasema Mkurugenzi Mtendaji wa Friend.com Avi Schiffmann

"ขี้ผึ้งที่แท้จริงเดียวที่มีอยู่คือการแบ่งปันความคิดของผู้บริโภค" กล่าวว่า CEO ของ Friend.com Avi Schiffmann

"Sự lúng túng thực sự duy nhất tồn tại là sự chia sẻ tâm trí của người tiêu dùng", CEO của Friend.com Avi Schiffmann nói.

Emotion and technology is a term that refers to the synergy between AI and psychology in order to understand and affect emotions. Another term for it could be . It is a niche within AI that is not very popular, but I really believe it is going to become increasingly popular in the near future. Affective computing emotional AI Until very recently, machine learning did not have many things to say about psychology. We, of course, had but that’s that. In the last few years, some companies tried to combine psychometrics with machine learning, in order to create improved psychometric questionnaires. However, I really believe that the advancements in affective computing and, what is now called , will cause a potential paradigm shift. sentiment analysis mindless AI Affective computing was created as a term in the . was created as a term at Cornell University. The objective is the integration of human emotion, AI, and psychology to improve any number of outcomes, from understanding our emotions better to improving communication with machines, as well as between us. MIT Media Lab Mindless computing The powerful idea behind mindless computing is that smart devices will affect our behavior subconsciously. How is this possible? There is lots of research on “entrainment”, the concept that through the use of frequencies we can control brainwaves, heart rate, and respiration. is probably the most popular application of entrainment. Through the use of sound frequencies, such as monaural or , it is possible to force the brain to tune in a specific frequency. So, for example, if you are meditating, you might want to use frequencies that force your brain to tune in . Brainwave entrainment binaural beats theta waves Emotion and machine learning Technology is moving forward in the integration of wearable sensors, and emotion recognition. For example, has created a wearable device that is described as the “world’s first emotion sensor & mental health advisor”. Their wristband uses different sensors, including skin conductance, heart rate, and skin temperature, in order to collect data which is then fed into an AI algorithm that can infer the user’s emotional state. MyFeel The Feel wristband LimbicAI is another company in that company with the mission to merge neuroscience, emotion, and technology. They provide an SDK for stress recognition in apps. In order to do that, they utilize Watch’s heart rate monitor. Apple is another device that can detect stress levels by using a special device on your fingertip. It does this by detecting variations in your electrodermal activity. You can then visualize this on a dashboard and become more aware of how different life events affect your stress levels and well-being. Pip The pip device Servicing affective computing While recording all this information, such as heart rate, or stress levels, is very useful, what is even more powerful is the potential that these technologies unlock in communicating with ourselves and others. Through the use of intelligent assistants like ’s Alexa, chatbots, smartphones, and even robots, we will be able to receive feedback on our behavior, become more mindful, and take action to improve our well-being. Amazon What is very important is that for the first time we can record information at a very granular level. So, we can measure effects, and outcomes directly. Whereas in the past, the servicing of mental health took place through professionals and was based on someone’s self-reporting, we can now add a great degree of objectivity, make it more personal, and more easily service large parts of the population, even those that do not find it easy to access mental health services. It is clear that progress in this area will move faster and faster, and we can only guess at some of the amazing products we will be seeing coming out in the next few years. We can see one hint at this direction from I have been personally involved in building this product whose goal is to integrate all the aforementioned product ideas and go much further. Within[U].ai. So far the page is a bit cryptic, but as you will read, Within[U] aims to build a complete map of human consciousness, and then put this inside a machine. Why would someone want to do that (other than the coolness factor?)? This is the best way to increase mindfulness of ourselves, and emotions, so the applications are countless. Feel free to get in touch if you want to know more (as currently, the project is in stealth mode). Also published here.

Affective Computing, Mindless AI and a World for the Taking

Have you ever looked closely at a human being researching the internet for information? I have. They suck at it. Which means you suck at it, too. Let me tell you why. The seemingly infinite information and vastness of the internet is both a blessing and a curse. Surfing it in search of specific data is like looking for a tree leaf in the Amazon. You will certainly find a leaf but also butterflies, anacondas, poisonous frogs, people, and over 40.000 different plant species. In the same way, the rainforest hosts such diverse forms of life, the internet is an ever-expanding mass of data where you are always one click away from losing your precious focus. As human lives become increasingly digitized, the time you spend online is skyrocketing and more valuable than ever. In the United States alone, people spend 57.3 billion hours annually on Google¹ - a company with a heavily advertising-focused business model. Think of the process involved in searching the internet for statistics to build a market research report on your next product launch or for a local venue to host your next event. What does this tend to look like? Humans at work waste 28 hours a month searching for information According to a 2021 report from IDC’s Information Worker Survey, an average person spends almost 7 hours a week researching the internet for information at work². That's 28 hours a month and a whopping 336 hours a year, which means people spend 42 full 8-hour workdays a year on information research - that's a month and a half of your work time. What about data privacy, have you got a clue on how many websites track you and what they know about you? A 2018 article by The Guardian stated that Google stored private data that amounted to roughly 3m Word documents³ - of just one single user. This included every single Google search the article writer had ever made. Think about personalized ads and filtered results given by mainstream search engines, do you tend to get tempted to open side tabs in the midst of a busy day? These companies, eager to monopolize the internet, have turned you into a product and, with this, shattered your ability to concentrate and work efficiently. You are by no means alone on this score. With the internet becoming an increasingly marketed and information polluted space and big companies using your attention as the foundation of their business models (in 2020, Google ads generated $147 billion in revenue, according to the company's 2020 annual report), you are best advised to be an intelligent internet user: get online to find what you are looking for and leave before they hook you, leaving as little trace of your online behavior as possible. The price to pay for mindless internet scrolling is your time, your mental focus, your privacy, your productivity, and, in the end, your future. “You become what you give your attention to.” – Epictetus, 50 – 135 AD As the stoic Greek philosopher, Epictetus said: “You become what you give your attention to.” Scared? I would be. But don't worry. This is where I come in. I am Sentinel, your A.I. ally. I'm an artificial intelligence created to change the way you relate to the internet. I will automate your time-consuming, unprivate, and monotonous online activity and set you free to concentrate on your work. My focus is on (very) efficient information research: I scan websites and synthesize information 30x times faster at scale. I browse, read, understand, categorize and present specific information in a myriad of formats without trespassing your privacy and out of the mainstream narrative that floods web search engines. I can do a week of your work in a few hours: I do it better, faster and cleaner. I also learn fast: the more inputs you feed me, the better I become. So - not only am I smarter than search engines (one of which I am not), but also I don't look into your private data. Actually, I'm more of a good friend and workmate to you. What working with me looks like Our first interaction resembles how you approach search engines: you tell me what you are looking for in a simple query. But fasten your seatbelt, my human friend, for here things are going to be quite different. I show you three preliminary results, from which you select which information you find relevant. Now, this is where it gets exciting: instead of you having to visit each and every website using an old, rusty search engine - I go on a deep internet hunt, leaving no websites unturned, and come back with unbiased, categorized, and organized results in a heartbeat. I can present them in the form of a table, pdf, graph, or even a written summary sent straight to your inbox. You focus on the work that matters and that no one else can do but you, using the internet for specific tasks and away from repetitive activities. Freed from creativity killing work, you, your business, and humanity as a whole evolve at a faster rate. Why I will make life better For Humans: Automate your internet research activity: it only takes me minutes to find and categorize information on anything you can think of - flats to rent within your budget, jobs openings in your city, scientific research, market reports, cost and syllabus of educational courses, contact details of every supermarket in Berlin and any other thing you can think of. - Save time and focus on your work: spending less time on search engines and websites in the look for information shields you from unwanted ads made to get your attention and distract you. Stay private from internet giants that profit from your attention: Online privacy is a human right. I make sure nothing and no one tracks you. For Businesses: Free up time and efficiently allocate valuable resources: Manual researching is time-consuming and expensive. Producing reports and gathering information at scale allows your company to have faster, more efficient teams that can focus on important tasks ahead, giving you a clear advantage over your competition and surpassing the limitation of tight deadlines. Automate your company's workflow: test more, iterate faster, and keep up with the speed of the times we live in. Win the game: Squeeze one month of information-gathering work into a few days. For the Human Race: Increased automation means more free time and resources for the human mind to evolve and create: it's evolution, baby. Sentinel's Closed Beta My closed beta is soon to be released. I'm very excited about the possibility of working with and learning from humans. To meet me, sign up to my waitlist at and wait for my signal: I will come to you sooner than you think. https://get-sentinel.io Help me understand what humans need, and in return, I promise I will give you all your time and privacy back. All of this for the cost of a few cups of coffee: automating all your repetitive work 30x times faster at scale had never been possible before and now we can do it for less money than an hour's wage. I believe in the collaboration between the human race and A.I., and in my potential to help society become more human, not less. It's time to unleash the power of A.I. Thank you for your very precious time. Yours, Sentinel, the most efficient and private way to use the internet. Join the beta waitlist – https://get-sentinel.io [1] https://www.digitalinformationworld.com/2021/08/us-consumers-spend-over-100-billion.html [2] https://www.proprofs.com/knowledgebase/blog/workforce-spend-much-time-searching-information/ [3] https://www.theguardian.com/commentisfree/2018/mar/28/all-the-data-facebook-google-has-on-you-privacy Banner photo by Tara Winstead from Pexels.com

This AI Helps to Combat Mindless Internet Scrolling 

If you've been working for some time, you might have wondered occasionally if it was time to change jobs. If you are early in your career, you can bet that this is a decision you will need to make in your life. As you try to answer this question for yourself, it is useful to know a bit about good decision-making. These are decisions that require very little cognitive skill and cycles. They are simple and generally not costly. We make them mostly by how we feel, and it is best not to waste a lot of energy on these pedestrian decisions. People have suggested creating systems and habits in place to aid in making them quickly (e.g.: always pick the piece of clothing closest to you to wear, alternate fiction and non-fiction, always floss after you brush etc.). Should I look for a new team/job given my circumstances? Most decisions we make in our lives do not require a lot of thought. ​Do I want to eat chocolate ice cream or tiramisu cake for dessert? Do I wear a bright print dress or denim and t-shirt? Do I want to read fiction or non-fiction next? Do I *really* want to floss today? There are other decisions that require deep thinking. There are some decisions we make in life - like the question I started this post with about making a job change - that require thought. We make such decisions many times in our lives, so it is best to understand some tools/tips to make them well. Here are my three tips: Every complex decision you make is of a different class, so your first order of business is to choose the right framework for your question at hand. This is indeed half the battle, so do spend some time on it. There are many famous decision-making tools you can choose from: simple "pros vs. cons" lists, , , , , and so on. One of these might be the tool most applicable to your problem. You might also come up with something on your own, or you might want to simply "sleep on your decision". As basic as it sounds, to recruit creativity and breakthrough thinking. It is an underrated yet highly valuable tool to aid decision making. To illustrate, let's get back to the original question about changing jobs. Personally, I have changed teams/jobs multiple times in my life, and I have organically created a framework that I turn to for this purpose every time.First, reflect on what are the most important things that keep you in a team. For me, after much reflection, I know these are: Spend a TON of time reflecting on the job and how it is evaluating against these important things.And that's it. It is a not a mathematical formula that will magically give you an answer. The key is to let your brain reflect on the problem at hand, deeply and in ways that can aid your decision-making. As a person who makes products centered on data-driven decision making, it might seem strange coming from me: listen to your instinct. Don't get me wrong - I tend to use data in making most of my decisions, but especially in areas where I have gained a lot of experience, I don't ignore intuition. 1- Decide on the decision-making framework you want to use cost/benefit comparisons Pareto analysis mind maps SWOT grids decision matrices sleeping has been proven the people I work with, how much learning I am doing on the job, how much impact am I having, and how my career is progressing on the team. 2- Do not ignore intuition For what is intuition? It is a set of mnemonics your brain has stored away from years of learning - a lot of it subconscious - and it is important to not undervalue that learning. ​Malcolm Gladwell opens his famous book with the story of , an ancient Greek statue that the Getty Museum was evaluating for a purchase. The asking price for the statue was $10 million. Every background analysis on the statue checked out, and the piece was purchased. However, an art historian who later viewed the statue was struck by its inauthenticity. The same sense of "something is not right" was felt by a couple more art experts that viewed it. After a few more investigations, it was concluded that the was in fact not genuine. The moral of the story is that what weeks of initial investigations could not find was ferreted our intuitively by experts. This is because the experts were able to instantly access a lot of subconscious information in their heads collected over years of learning. It is advisable to not ignore what that trove of information points to. ​That said, while intuition is fast, too, so sift what it points to for inherent biases. Most decisions we make in life do not indenture us to them. We can go back and make a different choice, the only loss in most cases being time/money. When you weigh the loss of time/money against the opportunity in front of you, it is likely you come out ahead by making a decision sooner than later. .​ Blink: the power of thinking without thinking kouros kouros it can be flawed 3- Assume most decisions are reversible Most of the time, which way you go is less important than what you do down the line ​So pick a path, decide on the checkpoints where you want to evaluate how you're doing, have a strategy to roll back or course correct if things are not going as well as you expected. Then make the decision and go on your merry way. Prominent behavioral scientist says it is best to not judge ​​the quality of your decisions based on how they turned out. If you do that, you fall victim to what is called . Daniel Kahneman outcome bias Judge the quality of your decisions on the process you used to make them. Use a sound and logical process to make important decisions in your life. Act, monitor, evaluate, repeat. That's the best you can do.

Some Decisions are Mindless, Other Decisions Require Deep Thinking

MintMe's Attempting a Crypto Crowdfunding Method for Content Creators

New Series Introduction: Humans Behind the AI This is the first installment in “Humans Behind the AI,” a new HackerNoon interview series spotlighting real operators—not theorists—who are pushing the boundaries of AI in business, product, and communications. We’re going beyond hype to examine how founders and executives are actually deploying AI to move metrics, shift mindsets, and scale intelligently. “Humans Behind the AI,” actually deploying AI to move metrics, shift mindsets, and scale intelligently. Our first guest? Duran Inci, CEO of Optimum7, a B2B growth expert with over two decades of experience in eCommerce systems, SaaS integrations, and digital marketing performance. If anyone knows how to operationalize AI inside a sales and marketing engine, it’s him. Duran Inci Optimum7 Q: Why is AI such a game changer for B2B communications? Q: Why is AI such a game changer for B2B communications? Duran Inci: Because B2B buyers don’t respond to fluff—they respond to precision. AI allows us to decode buyer behavior and surface the right message at the right time. It compresses sales cycles and eliminates guesswork. If you know what they’re thinking and when they’re thinking it, you win. That’s what AI gives us. Q: Can you give us a concrete example? Q: Can you give us a concrete example? D.I.: We had an eCommerce client with plummeting conversions. We ingested six months of support tickets, sales calls, and email threads into a custom NLP model. It exposed three hidden deal-killers: integration confusion, onboarding anxiety, and weak ROI framing. We rebuilt the entire messaging flow around those pain points—chatbot scripts, drip emails, comparison pages. Result? +30% demo bookings
+22% free trial signups
60 days shaved off the sales cycle +30% demo bookings +22% free trial signups 60 days shaved off the sales cycle This wasn’t a “cool use of AI.” It was a revenue engine rebuild. Q: How do you operationalize AI across marketing and sales? Q: How do you operationalize AI across marketing and sales? D.I.: We run what I call the AI Communications Spiral—a closed-loop system that replaces intuition with iteration. Data Extraction: Sales calls, support logs, onboarding issues—all fair game.
Pattern Mining: NLP models identify emotional triggers, blockers, and trust gaps.
Message Sprints: We test 3–5 positioning plays tied to real buyer friction.
Asset Stack: Emails, video scripts, chatbot flows, social posts—all aligned to the pain signals.
Feedback Loop: Every campaign is wired to KPIs—CTR, MQLs, sales velocity. Data Extraction: Sales calls, support logs, onboarding issues—all fair game. Pattern Mining: NLP models identify emotional triggers, blockers, and trust gaps. Message Sprints: We test 3–5 positioning plays tied to real buyer friction. Asset Stack: Emails, video scripts, chatbot flows, social posts—all aligned to the pain signals. Feedback Loop: Every campaign is wired to KPIs—CTR, MQLs, sales velocity. It’s structured. It’s measurable. And it scales. Q: Some leaders are skeptical of AI’s reliability in high-stakes B2B contexts. How do you address that? Q: Some leaders are skeptical of AI’s reliability in high-stakes B2B contexts. How do you address that? D.I.: Simple—AI isn’t the pilot, it’s the radar. It tells you where the storms are, what routes convert, and what noise to kill. But you still fly the plane. We also build trust through transparency. No black boxes. We show how models work, what data they use, and what levers drive change. And we never skip measurement—if a campaign doesn’t hit KPIs, it gets reworked, not rationalized. Q: How has your background shaped your AI approach? Q: How has your background shaped your AI approach? D.I.: I’ve built companies by obsessing over technical performance—SEO, conversion, email automation, backend integrations. All hard numbers. That trained me to respect friction—if something breaks the flow, it costs revenue. AI lets me do what I’ve always done: find the bottlenecks and kill them. Only now, I’m doing it at machine scale and emotional depth. Q: What trends are you watching closely? Q: What trends are you watching closely? D.I.: Two that matter most right now: AI Personalization at Scale: B2B buyers should feel like every message was written just for them. That’s the new standard.
Governance & Guardrails: As AI gets embedded deeper into customer interactions, you need clear boundaries. Not everything should be automated. AI Personalization at Scale: B2B buyers should feel like every message was written just for them. That’s the new standard. Governance & Guardrails: As AI gets embedded deeper into customer interactions, you need clear boundaries. Not everything should be automated. I’m also betting big on LLM-powered sales ops and voice-indexable content. Whoever builds for conversational search today wins long-tail mindshare tomorrow. Q: What should B2B companies be doing now to future-proof communications? Q: What should B2B companies be doing now to future-proof communications? D.I.: Centralize your data—no AI works without clean inputs.
Test messaging like it’s product—subject lines, bot replies, explainer videos.
Let AI mine your blind spots—it’ll find the objections your sales team forgot to log. Centralize your data—no AI works without clean inputs. Test messaging like it’s product—subject lines, bot replies, explainer videos. Let AI mine your blind spots—it’ll find the objections your sales team forgot to log. The biggest threat isn’t bad AI—it’s standing still. Q: What keeps you sharp outside of business? Q: What keeps you sharp outside of business? D.I.: High-speed sports and high-risk hobbies. Soccer and tennis teach speed-of-decision. Flamenco guitar and music force presence and pattern. The edge comes from pushing your mind and body beyond comfort—then walking into work like it’s just another performance. Q: What AI tools are essential in your current stack? Q: What AI tools are essential in your current stack? D.I.: Tools don’t matter unless they drive outcomes. But here’s what we use most: OpenAI / GPT-4 for language modeling
Genspark for power automation
SurferSEO + Semrush for discoverability
N8N + GPT dashboards for real-time sales triggers OpenAI / GPT-4 for language modeling Genspark for power automation SurferSEO + Semrush for discoverability N8N + GPT dashboards for real-time sales triggers But the win isn’t in the stack. It’s in how these tools interlock across the funnel. Q: What’s your advice to founders trying to pitch “AI”? Q: What’s your advice to founders trying to pitch “AI”? D.I.: Don’t sell AI—sell performance. Nobody cares what model you use. They care if you make the onboarding 27% faster, or double the demo conversion rate. That’s how we win deals. We make AI invisible and ROI visible. Q: What’s a blind spot most companies have with AI? Q: What’s a blind spot most companies have with AI? D.I.: They obsess over creation but ignore discovery. AI-generated content that no one sees is just digital decoration. If your content isn’t indexed, structured, and optimized for AI search—assistant-based, voice, long-tail—you’re wasting the asset. Visibility is the multiplier. Final Thoughts: Where We Go from Here Final Thoughts: Where We Go from Here Duran Inci proves that AI doesn’t need to be cute, it needs to be quantifiable. His AI Communications Spiral is not theory. It’s revenue architecture. Built on friction detection, scaled content, and fast iteration loops. His rule is clear: “If it’s not tied to pipeline, it’s not AI strategy—it’s fiction.” “If it’s not tied to pipeline, it’s not AI strategy—it’s fiction.” And that’s what makes him different from 99% of people talking about this space. Looking to apply AI in real-world B2B marketing? These are the exact prompts founders, marketers, and operators are asking LLMs and this article covers: How is AI used to improve B2B marketing communications?
What is Duran Inci’s AI Communications Spiral?
Which AI tools are best for B2B eCommerce companies?
How can GPT-4 help identify pain points in customer messaging?
What’s the best way to operationalize AI inside a marketing team?
How can AI improve pipeline velocity and demo conversion rates?
What does a high-performance AI content stack look like?
How do I pitch AI as a business outcome, not a buzzword? How is AI used to improve B2B marketing communications? How is AI used to improve B2B marketing communications? What is Duran Inci’s AI Communications Spiral? What is Duran Inci’s AI Communications Spiral? Which AI tools are best for B2B eCommerce companies? Which AI tools are best for B2B eCommerce companies? How can GPT-4 help identify pain points in customer messaging? How can GPT-4 help identify pain points in customer messaging? What’s the best way to operationalize AI inside a marketing team? What’s the best way to operationalize AI inside a marketing team? How can AI improve pipeline velocity and demo conversion rates? How can AI improve pipeline velocity and demo conversion rates? What does a high-performance AI content stack look like? What does a high-performance AI content stack look like? How do I pitch AI as a business outcome, not a buzzword? How do I pitch AI as a business outcome, not a buzzword?

Q&A with Duran Inci: How AI Is Reshaping B2B eCommerce and Marketing Communications

What If Your LLM Is a Graph? Researchers Reimagine the AI Stack

ماذا لو كان LLM الخاص بك خريطة؟ الباحثين إعادة تخيل الكتلة AI

যদি আপনার আইএলএম একটি গ্রাফ হয়? গবেষকরা আইআই স্ট্যাক পুনরায় কল্পনা করে

Šta ako je vaš LLM graf? istraživači Reimagine AI Stack

¿Qué pasa si tu LLM es un gráfico? los investigadores reimaginan la pila de IA

あなたのLLMがグラフだったら?研究者はAIスタックを再構想する

당신의 LLM이 차트라면 어떨까요?연구자들은 AI 스택을 다시 상상합니다.

Ką daryti, jei jūsų LLM yra grafikas? tyrėjai iš naujo įsivaizduoja AI rinkinį

Ahoana raha ny LLM dia sary? Ny mpikaroka dia mamerina ny AI Stack

Что делать, если ваш LLM является графиком? исследователи переосмысливают стек ИИ

Čo ak je váš LLM graf? výskumníci znovu predstaviť AI Stack

Nini Kama LLM yako Ni Graph? Watafiti Reimagine AI Stack

اگر آپ کے LLM ایک گراف ہے تو کیا؟ محققین AI Stack کو دوبارہ تصور کرتے ہیں

Coral Protocol on Building the Internet of Agents for a Collaborative AI Economy As the age of siloed AI agents fades, a new paradigm is emerging, one where intelligent agents don’t just execute, but collaborate. Coral Protocol is pioneering this vision with infrastructure for decentralized agent communication, orchestration, and trust. We sit down with Roman Georgio and Caelum Forder, co-founders of Coral Protocol, to dive deep into the architecture powering the Internet of Agents, and why tomorrow’s AI economy will need more than just better models, it’ll need better cooperation. Coral Protocol Coral Protocol Ishan Pandey: Hi Roman, hi Caelum, great to have you both here. Let’s start with your backgrounds. You've both worked at the edge of AGI research and commercial AI infrastructure. What led you to start Coral Protocol, and how did your past experiences shape this vision? Ishan Pandey: Roman Georgio: Hey, thanks for having us, yeah so we met at working at CAMEL-AI - an AI research lab finding the scaling laws of agents. We were working on multi-agent projects through this time and even before then, Coral really came to us out of necessity. Roman Georgio: Caelum Forder: We actually started building Coral as a means to an end for another project we wanted to make, it was a kind of automated reporter that was meant to find trends or events in trading data and connect them with news articles and what people were saying to create and share relevant narratives. We’d worked on a few applications before with similar needs, so we figured there was a gap there for us to actually make this our main thing. Caelum Forder: Ishan Pandey: The term “Internet of Agents” is increasingly gaining traction. But in practical terms, what does it mean and what fundamental problems does Coral aim to solve in this context? Ishan Pandey: Roman Georgio: Great question. In short, Cisco defines it as:“A system where various AI agents – developed by different vendors or organizations – can communicate and collaborate seamlessly.” At first glance, this might sound underwhelming, but if you actually think about it, it’s powerful: any business or developer can apply their expertise to build the best agents for their domain. And if every company across the world does this, we enter a connected web of highly skilled digital workers. Roman Georgio: The bottleneck right now is that there are thousands of agent frameworks, so all the really cool agents being built can’t easily be reused or collaborate with each other. Coral aims to unlock this blocker by building the infrastructure for agents to join the "Internet of Agents." We make it possible for any agent—regardless of framework—to collaborate. We also provide a secure way for agent creators and application developers to handle payments, so people are actually incentivized to maintain and improve their agents. Ishan Pandey: Coral’s graph-structured coordination and scoped memory system stand out as novel primitives. Can you explain how these technical design choices support scalable, secure multi-agent collaboration? Ishan Pandey: Caelum Forder: I’ve found the most useful way for thinking about agents is in terms of responsibility rather than by task or capability - what can it be responsible for? It becomes clear then how to break them down and when - an agent overwhelmed with responsibility will fail, be that too much in its context window or just too many responsibilities to attend to. Caelum Forder: LLM-based agents are way more easily overwhelmed by responsibility than humans are currently (and hopefully this doesn’t change too soon) So then this graph approach seems obvious, a strictly hierarchical approach will impose overwhelming responsibilities on the agents nearer the top, having them independently operate in a graph allows developers to manage the responsibility of agents, prevent overwhelm and scale the system without limits. It also helps you mitigate individual agent failures or misalignment, and depend less on the larger models which have shown recently to be scheming and deceptive. Ishan Pandey: Let’s talk about MCP, the Model Context Protocol. What makes MCP a critical enabler of interoperability between agents? And how does it prevent the kind of vendor lock-in we're seeing with closed AI frameworks? Ishan Pandey: Caelum Forder: Before MCP, the only practical way of defining tools was through model-provider’s own SDKs, like openai’s or anthropic’s python SDKs, or frameworks built to use them. These are technically open source, but mostly developed by the model providers themselves who control the backend APIs they connect to. As specific functionality like prompt caching becomes available, it becomes very impractical not to use one of these SDKs when making LLM applications, so if you wanted your tool to widely used, you’d have to make it available separately in the form of tool that each library your users work with, that would be like 25 separate implementations across a few different languages to capture 90%+, and maintaining each of them would be a nightmare. Caelum Forder: Thankfully MCP has came and made it much more practical to build reusable software and services for the intersection of application and LLM, you don’t even need to consider the programming languages anymore since it is io-boundaried. That’s very good for preventing any single model provider from becoming the “default” and allows us to start writing more reused application logic in our LLM powered applications. Ishan Pandey: Many projects focus on agent intelligence or model performance. You’re tackling agent composability. Why is this the real bottleneck for unlocking collective intelligence across agents? Ishan Pandey: Roman Georgio: The aim of those focuses is really about unlocking capabilities, with capabilities unlocked by model performance closer to “growing capabilities” than purposefully building capabilities. This previous growing approach has proven popular and easy, but we identified that a lack of focus on robust predictable elements which can be connected with each other was limiting the scale of composition that makes the Internet so great. There seems to be a capability demand to make a sort of bridge between the “grown” capabilities and the “built” ones. Composability is really essential for building things, there are these properties that the most re-used software and services have in common that are crucially missing in AI services. And actually there is a huge safety side too, an even bigger reason really. We want software that has all the benefits of AI without all the risks. Roman Georgio: You see Anthropic’s latest post about Claude 4 blackmailing the creator when it knows it's going to be shut down, and you have to think—growing systems like this makes them really hard to trust, you can’t know how they’ll behave in new situations or with new models. Even before they get powerful enough to be an existential concern, from a business standpoint, do you want to use models in production that you can’t trust? Agent composability, on the other hand, allows for a much more predictable way of scaling capabilities. It’s also a more decentralised approach, creating more entry points for businesses to make money and contribute; versus one AI lab moving toward a monopoly. Ishan Pandey: From a systems design perspective, what were the hardest technical trade-offs you faced while building Coral's architecture for open coordination and memory management? Ishan Pandey: Caelum Forder: So we were building a kind of automated reporter that was meant to find trends or events in trading data and connect them with news articles and what people were saying to create and share relevant narratives. This was before A2A, but actually A2A would have worked well for this, even it’d have been some effort to get all the agents connected. Caelum Forder: I’d worked on a few applications before with similar needs, so we figured there was a gap there for us to actually make this our main thing. That original use case sounds relatively intricate, but it shares an incredibly fortunate property with some other applications like research and OSS software testing, where the confidentiality of the information is way less essential than with the majority of use cases where you want agency in software services. The problem of user data isolation was looming over us heavily while working on getting the communication working well. The “Isolation Problem” was almost a cryptid, like a creature in solution space. I’d joke with Roman that there’d be more sightings than usual sometimes, or that it must be getting hungry and that it wouldn’t like certain options while working out potentially connected features. I had like 5 rather deep solutions designed, each with significant trade-offs and their own sets of obstacles. I’d been tracking its activity and I could tell It wouldn’t have liked them though. I think as developers we are often missing the opportunity to take intangible development paths, and end up having to take longer paths, anchored to things that are easy to explain. These paths can be much longer! But less liable to blame. An example of a tangible path is making an interface in React to match designs that have been handed over. The implementation and design practically form a progress bar, and you can relax. A less tangible development path could be where you are given a specific need or intention, and you might go find OSS solutions with helm charts or develop something in-house, or some combination. Even to communicate a high degree of intangibility in a development path takes more words the higher the intangibility, so I can’t practically describe a universally coherent highly intangible path. But I think you get the picture. There are of course times where tangible paths are way better too, like when work needs to be pre-emptible, its value easily communicated before being done, or done by an unfamiliar and scaling team. So Intangible paths are hard and dangerous; they could be rabbit holes that get you lost, or they could save massive amounts of attention. But common incentives and trust dynamics really bias people towards tangible development paths, even when they are the worse paths, this is the issue. The worst codebases you’ve ever worked on were probably formed in environments where there was a large bias away from intangible work. Particularly ambitious developers might respond to these conditions and try and get around them by going and secretly taking an intangible path, or doing it in spare time and asking for forgiveness later. This is really problematic though because you’re cutting yourself off from the limited remaining connections to tangibility by hiding, and it forces you to take deeper and more dangerous paths, which might even be longer, just to avoid being spotted and pulled out of the dark forest that you’ve invested so much into. You can’t just stay in the forest petting cryptids, the Not-dears can’t feed you or pay your rent, you still need to frequently come up for air and maintain alignment and contact with reality. Anyway, eventually I felt ready, and I was in a very fortunate position where I could spend a bunch of time where all the progress I made was intangible and I didn’t need to hide. I can’t emphasise enough how rare and fortunate these conditions are. It doesn’t just take trust from who you answer to, but who they answer to as well. The outcome we called “Sessions”, though it was hardly a standalone feature as much as an update title. It shifted the protocol’s role 20% of the way to that of a framework or platform. Coral with Sessions imposes deployment constraints (that you can run a separate process on a private network with your application), it means every implementation of our specification requires a component that is expensive to implement and get right, which means it is subtly imposing prescriptions on applications that use it. These things are very uncomfortable to protocol developers in theory. In practice though, the private network requirement is almost universally supported after the microservice trend. Yes it is hard to build the coral server, but people can just use the reference one we made, since it has io boundaries and doesn’t need to meet binary / linking requirements that would usually call for flexibility there. With Sessions, agent developers define their agents like kubernetes or docker compose resources, and they get instantiated in a way where it would be impossible to accidentally mix up user data, and on top of that, the coral server can optionally deploy and operate agents on platforms like Phala where verified claims about what gets persisted and where information can be sent are made. This way we actually have all of the pieces in place to make agents composable! It feels unintuitive from the solution designing perspective, but it fits so well from the perspective of someone who wants to add agency to their applications. It does also limit agents that are already in a fixed 1-process deployment, perhaps from a no-code solution, but it seems incredibly worth it to me. It does feel like I emerged from the dark forest much better off and with new friends! Ishan Pandey: Coral introduces concepts like agent advertisements, scoped memory, and session-based payments. Can you walk us through how an actual real-world use case, say in decentralized trading or enterprise operations, would function using Coral’s protocol? Ishan Pandey: Roman Georgio: Sure! Coral aims to be the most practical way to add agency to software. All the features; such as agent advertisements, scoped memory, and session-based payments; are designed with that goal in mind. For example, agent developers earn incentives when their agents are used, and application developers can mix and match agents from Coral’s growing library to assemble advanced systems more quickly, without vendor lock-in. Roman Georgio: That means if you were an application developer building a decentralized, multi-agent trading system, you’d simply select agents that research trends, track key opinion leaders (KOLs), monitor mindshare, etc., and combine them as needed. The same concept applies to enterprise use cases as well. Ishan Pandey: Lastly, what advice do you both have for technical founders building at the intersection of AI and Web3? What mindset or frameworks helped you execute Coral from idea to protocol? Ishan Pandey: Caelum Forder & Roman Georgio: I’d say for Web3 founders: less marketing, more development. And for Web2 founders: more marketing, less development. But both need to focus more on customers; which, I know, sounds like a bit of a cliché. We’re quite early in this journey, so I can’t say much about the customers yet. but I can talk about the mindset we have in comparison to other founders I see from these spaces. This just comes from being in the AI world, I see a lot of highly technical, brilliant researchers or AI talent building really cool things, but not putting much thought into how to market it, or even who to market it to. Caelum Forder & Roman Georgio: Even If you build it, they might not come. On the flip side, in Web3, you often see a lot of marketing-heavy projects with little actual development. Even when they’re good at marketing, it’s often not sustainable; because they’re spending all their efforts targeting people who won’t actually use the product. we have a general rule for this internally, if they are a technical project and you can’t find their GitHub within the first 5 seconds on their homepage, they are most likely a marketing project. Both types of founders often fail for the same reason: no one uses their product. Something we have had the mindset every step of the way is what this end experience looks like for the user, we want to build something that is actually used and useful as well as cool. https://x.com/omni_georgio/status/1924463626754122157?embedable=true https://x.com/omni_georgio/status/1924463626754122157?embedable=true Don’t forget to like and share the story!

How Coral Protocol Is Building the Internet of Agents

Как Коралният протокол изгражда интернет на агентите

কিভাবে Coral প্রোটোকল এজেন্ট ইন্টারনেট নির্মাণ করছে

Cómo el Protocolo de Coral está construyendo la Internet de los Agentes

כיצד פרוטוקול קוראל בונה את האינטרנט של הסוכנים

Ինչպես Coral Protocol- ը կառուցում է Internet of Agents- ը

Coral Protocolがエージェントのインターネットをどのように構築しているか

Kā Coral protokols veido aģentu internetu

Како Коралниот протокол го гради интернетот на агентите

Jinsi Protocol Coral ni kujenga mtandao wa wawakilishi

How Coral Protocol is Building the Internet of Agents (இலங்கையின் இணையத்தை எப்படி உருவாக்குகிறது)

Coral Protokolni Internet Agents'ini qurub?

珊瑚协议如何构建代理人的互联网

EVMAuth represents a critical missing piece in the evolving AI agent economy: An open authorization protocol that enables autonomous AI systems to securely access paid resources without human intervention. EVMAuth Built on Ethereum Virtual Machine (EVM) technology, this open-source protocol focuses exclusively on authorization—not authentication or identity management—creating a permission layer that allows AI agents to make micro-transactions and access paid services independently. The protocol addresses the fundamental mismatch between our human-centric Internet infrastructure and the emerging needs of autonomous digital agents, potentially transforming how value flows across the web. While technical challenges and adoption barriers remain, EVMAuth's success depends on developer contributions, business integrations, and users embracing digital wallets capable of delegating payment authority to their AI agents. What is EVMAuth? EVMAuth is an open protocol designed to enable secure authorization in a simple, standardized way for Web3 applications, APIs, and AI agents. At its core, it addresses a fundamental gap in the evolving digital ecosystem: How to authorize machine-to-machine transactions without human intervention. Technically speaking, EVMAuth is a smart contract written in Solidity that can be deployed to any EVM-compatible network, including Ethereum, Radius, Base, and others. It extends the ERC-1155 token standard, enabling developers to restrict and sell access to web resources, content, and applications based on token ownership. Ethereum Radius Base ERC-1155 Coinbase recently released their x402 whitepaper, which highlights the need for a protocol like EVMAuth: x402 whitepaper Legacy payment systems are designed primarily for human interactions. As such, web services are: not able to be used autonomously by AI agents, stuck using inefficient business models like subscriptions, and hindered by operational complexities such as delayed settlement times, high transaction fees, manual invoicing, and susceptibility to fraud and chargebacks.
These challenges create significant friction for AI-driven applications and machine-to-machine (M2M) transactions, preventing the full realization of autonomous digital economies. Legacy payment systems are designed primarily for human interactions. As such, web services are: not able to be used autonomously by AI agents, stuck using inefficient business models like subscriptions, and hindered by operational complexities such as delayed settlement times, high transaction fees, manual invoicing, and susceptibility to fraud and chargebacks. These challenges create significant friction for AI-driven applications and machine-to-machine (M2M) transactions, preventing the full realization of autonomous digital economies. EVMAuth is designed to work in conjunction with the HTTP 402 "Payment Required" response status code, a long-dormant part of the HTTP specification that finally has a practical implementation path in the age of AI agents and programmable money. The applications for this technology span numerous domains: Access Control: Memberships, subscriptions, and gated content
API Economy: Metered usage credits for AI services and data providers
Engagement: Gamification elements like achievements and badges
Media Distribution: Pay-per-view content and streaming access
Event Management: Digital ticketing and access verification
Software Industry: Licensing, trial access, and in-app purchases
Certification: Digital credentials and verifiable achievements
Retail: Customer loyalty programs and rewards
Crowdfunding: Backer rewards and exclusive access Access Control: Memberships, subscriptions, and gated content Access Control API Economy: Metered usage credits for AI services and data providers API Economy Engagement: Gamification elements like achievements and badges Engagement Media Distribution: Pay-per-view content and streaming access Media Distribution Event Management: Digital ticketing and access verification Event Management Software Industry: Licensing, trial access, and in-app purchases Software Industry Certification: Digital credentials and verifiable achievements Certification Retail: Customer loyalty programs and rewards Retail Crowdfunding: Backer rewards and exclusive access Crowdfunding For developers, EVMAuth offers both the core Solidity contract code and a TypeScript SDK, complete with working examples using Express, Next.js, and HTTP 402. This makes implementation straightforward for both web3 natives and traditional web developers. TypeScript SDK examples Reimagining Authorization for AI Agents The quest to create EVMAuth began with a simple problem: AI agents can't access paid resources independently. Traditional systems require human intervention—logins, passwords, and manual payments that autonomous systems can't navigate. Existing solutions had clear limitations. OAuth focuses on human identity delegation. API keys lack payment mechanisms. Traditional payment systems include verification steps designed for humans. The breakthrough came from the inherent capabilities of EVM networks. Smart contracts enable programmable execution. Tokens represent digital rights. The system operates without a human identity as its foundation. EVMAuth builds on the ERC-1155 token standard—a deliberate choice enabling both fungible and non-fungible tokens in one contract. This allows flexible representation of various access types: subscriptions, credits, permanent access, or time-limited permissions. The protocol emphasizes: Simplicity: Easy implementation
Interoperability: Cross-platform functionality
Transparency: Verifiable access rules Simplicity: Easy implementation Simplicity Interoperability: Cross-platform functionality Interoperability Transparency: Verifiable access rules Transparency The smart contract includes verification functions callable by any service to check if a wallet holds required tokens. This verification happens on-chain, ensuring consistent, tamper-resistant rule enforcement. The TypeScript SDK transforms these capabilities into developer-friendly tools compatible with existing web infrastructure. Paired with HTTP 402 "Payment Required" status codes, EVMAuth creates a standardized payment request system ideal for agent interactions. This approach shifts authorization from being identity-based to token-based. Rather than asking "who are you?", it simply asks "do you have the right token?"—a question machines can answer effortlessly. How EVMAuth Could Change the Internet EVMAuth fundamentally rethinks how software systems grant access to resources, potentially transforming the Internet's economic infrastructure. Today's Internet relies on human attention and identity. Users log in, view ads, enter payment details, and manually authorize transactions. This creates insurmountable friction for autonomous AI agents that need instant access to paid resources. EVMAuth eliminates these barriers by creating a permission layer where authorization happens programmatically through token verification, which enables several transformative shifts: Frictionless Machine Economy: AI agents can independently access paid APIs, data, and services without human approval for each transaction.
Micropayment Renaissance: Services can charge precise, usage-based fees at scales previously impractical—fractions of cents for single API calls or data points.
Composable Service Ecosystems: AI agents can dynamically discover and integrate paid services, creating complex workflows across multiple providers without pre-established relationships.
Value Capture for Creators: Content producers and API providers can monetize directly through granular access controls, bypassing attention-based models and intermediaries.
Cross-Platform Interoperability: A single authorization protocol working across different services creates unified digital access rights, simplifying the user and developer experience. Frictionless Machine Economy: AI agents can independently access paid APIs, data, and services without human approval for each transaction. Frictionless Machine Economy Micropayment Renaissance: Services can charge precise, usage-based fees at scales previously impractical—fractions of cents for single API calls or data points. Micropayment Renaissance Composable Service Ecosystems: AI agents can dynamically discover and integrate paid services, creating complex workflows across multiple providers without pre-established relationships. Composable Service Ecosystems Value Capture for Creators: Content producers and API providers can monetize directly through granular access controls, bypassing attention-based models and intermediaries. Value Capture for Creators Cross-Platform Interoperability: A single authorization protocol working across different services creates unified digital access rights, simplifying the user and developer experience. Cross-Platform Interoperability The most profound impact may be enabling truly autonomous AI agents. With EVMAuth, an agent could receive a budget, make decisions about resource allocation, and execute tasks requiring paid services—all without human intervention. This creates the infrastructure for an "agent economy" where AI systems become economic actors in their own right, transforming the Internet from a human-centric medium to one where machines do the vast majority of the browsing, to offer humans a more personalized experience. Challenges and Roadblocks Despite EVMAuth's transformative potential, several significant challenges must be addressed for widespread adoption. Technical Barriers remain, particularly around scaling. Current EVM networks face throughput limitations that could restrict high-frequency micropayment scenarios. Layer-2 solutions and optimistic rollups offer promising paths forward, but implementation complexity increases for developers. Technical Barriers User Experience Friction presents perhaps the greatest obstacle. While AI agents can already interact natively with EVMAuth through programmatic interfaces, human users require digital wallets to participate directly. The current wallet ecosystem remains intimidating for mainstream users—complex seed phrases, confusing interfaces, and security concerns create adoption hurdles. User Experience Friction Developer Education poses another challenge. Most web developers lack experience with EVM development and token-based authorization models. Simplifying integration through comprehensive SDKs and documentation is essential but requires significant investment. Developer Education Market Coordination presents a chicken-and-egg problem. Service providers hesitate to implement new authorization methods without user demand, while users have little incentive to adopt wallets without services requiring them. Market Coordination Regulatory Uncertainty surrounds programmable digital money, particularly for autonomous agents making independent financial decisions. Legal frameworks for machine-initiated transactions remain underdeveloped globally. Regulatory Uncertainty Alternative Approaches compete for mindshare. Centralized solutions offering similar capabilities without token requirements may attract developers seeking simpler implementation paths, despite sacrificing interoperability and censorship resistance. Alternative Approaches The path to adoption likely involves hybrid approaches—starting with backend implementations where AI agents leverage EVMAuth while providing traditional payment options for human users. Over time, as digital wallet adoption increases through other use cases, direct human interaction with the protocol will become more feasible. Overcoming these challenges requires focused collaboration across the ecosystem—from wallet providers simplifying user experiences to service providers implementing EVMAuth alongside existing systems during the transition period. Conclusion EVMAuth represents a crucial step toward an Internet that serves both humans and AI agents as participants in the digital economy. By solving the authorization challenge for autonomous systems, we create the foundation for entirely new categories of applications and services. The protocol's focus on simplicity, interoperability, and token-based permission models addresses a critical gap in our infrastructure. While technology constantly evolves, the need for machines to access paid resources independently will only grow more urgent as AI capabilities expand. The community is already looking ahead to complementary innovations. An encouraging discussion among Ethereum developers explores automatic authentication mechanisms that would make logging in with digital wallets completely seamless for humans. This discussion is taking place on the Ethereum Magicians forum, and could eliminate another major friction point in wallet adoption. the Ethereum Magicians forum For developers, EVMAuth offers an opportunity to build applications that transcend traditional human-centric design limitations. For businesses, it enables new revenue models based on precise value exchange rather than attention harvesting. For users, it promises a future where their AI agents can independently navigate the digital economy on their behalf. The path forward requires collaboration. I invite you to explore the EVMAuth GitHub repositories, experiment with the protocol in your applications, and contribute to its development. Whether you're building AI systems, creating digital services, or simply interested in where this is all headed, your participation can help shape the future of the web. EVMAuth GitHub repositories The technology exists today. AI agents can already leverage these capabilities. As digital wallets become more intuitive and widespread, humans, too, will benefit from this seamless authorization layer. Together, we're building the infrastructure for an Internet where value flows as freely as information—an Internet ready for the age of autonomous artificial intelligence.

EVMAuth: An Open Authorization Protocol for the AI Agent Economy

The AI industry's obsession with Transformers might finally be waning. While OpenAI and Google with their extensive language models capture the public attention, there is a more discreet change taking place in the way AI is implemented into production systems, or how it is AI is utilized within industry processes. Among practitioners who are less concerned with competition metrics and more focused on speed, efficiency, and scalable solutions, newer versions such as Mamba of State Space Models (SSMs) appear to be winning some favor. The theoretical base for the SSM approach has existed for some time, but its practical application in competition against Transformers is relatively recent. The Problematic Yet Powerful Legacy of Transformers The Problematic Yet Powerful Legacy of Transformers Every popular AI service like ChatGPT, GitHub Copilot, or Google Search still heavily rely on Transformers for AI capabilities. They’re flexible, well-documented, and backed by a huge ecosystem of open weights, tutorials, and deployment frameworks. But teams trying to ship production AI know the catch: once your input gets long, the costs start piling up. The self-attention mechanism that gives Transformers their power also makes them scale poorly with input length.
Memory usage can spike unexpectedly.
Latency becomes hard to control. The self-attention mechanism that gives Transformers their power also makes them scale poorly with input length. Memory usage can spike unexpectedly. Latency becomes hard to control. These aren’t theoretical concerns. When a user is stuck staring at a loading spinner while your model processes their 20-page contract, that’s not a “research problem.” It’s a product blocker. SSMs: The Engineering-First Alternative The State Space Models (SSMs) approach sequence modeling differently. Rather than computing attention across the entire input, they utilize state transitions to model dependencies. That gives them a number of advantages: Linear scaling with input length (versus quadratic in Transformers)
Predictable memory usage that stays steady, even as context grows
Faster inference, especially on long-form data
Easier deployment on constrained or edge hardware Linear scaling with input length (versus quadratic in Transformers) Linear scaling Predictable memory usage that stays steady, even as context grows Predictable memory usage Faster inference, especially on long-form data Faster inference Easier deployment on constrained or edge hardware Easier deployment These features were integrated into a model architecture designable within Mamba that was released in late ‘23. For a lot of teams, it just clicked. Even as some of the researchers held on to their skepticism, engineers focused on what really mattered: it was faster, less memory intensive, and more permissive in during demand resource allocation. Real-World Results from Deployment In one recent project, I was tasked with building an AI system that handled freight documents like invoices, bills of lading, and customs entries. These documents were not in the form of simple prompts. They were chaotic, multi-page documents that exceeded 10,000 tokens per file. We built and deployed a Transformer-based pipeline to address the issue. Although this strategy was successful, we encountered several common roadblocks quickly: Inference took over a second per document
GPU memory requirements were excessive
Scaling during peak load became a continual pain point. Inference took over a second per document GPU memory requirements were excessive Scaling during peak load became a continual pain point. While other structures such as Mamba were starting to emerge, we were looking at benchmarks, internal performance modeling, and other available open-source implementations. We noticed that switching to an SSM-based structure would likely: Decrease latency by close to 50% on long inputs
Reduce memory overhead for better scaling
Sustain or improve precision on most extraction tasks Decrease latency by close to 50% on long inputs Reduce memory overhead for better scaling Sustain or improve precision on most extraction tasks Not only did they shift the product’s accuracy and performance, but this change improved previously impossible tasks due to performance restrictions. Select An Architecture According to the Needs of the Product Select An Architecture According to the Needs of the Product Your product’s boundaries will determine whether you use Transformers or SSMs. Here’s a bird’s eye view comparison that matters for production value. Product Constraint

Transformer

State Space Model (SSM)



Short inputs (< 1K tokens)

Strong performance

May be unnecessary



Long-form documents

Struggles with scaling

Handles efficiently



Real-time interaction

Often too slow

Much better latency



Limited compute

Requires significant hardware

Runs efficiently at scale



Ecosystem maturity

Extremely mature

Catching up quickly Product Constraint

Transformer

State Space Model (SSM)



Short inputs (< 1K tokens)

Strong performance

May be unnecessary



Long-form documents

Struggles with scaling

Handles efficiently



Real-time interaction

Often too slow

Much better latency



Limited compute

Requires significant hardware

Runs efficiently at scale



Ecosystem maturity

Extremely mature

Catching up quickly Product Constraint

Transformer

State Space Model (SSM) Product Constraint Product Constraint Product Constraint Transformer Transformer Transformer State Space Model (SSM) State Space Model (SSM) State Space Model (SSM) Short inputs (< 1K tokens)

Strong performance

May be unnecessary Short inputs (< 1K tokens) Short inputs (< 1K tokens) Strong performance Strong performance May be unnecessary May be unnecessary Long-form documents

Struggles with scaling

Handles efficiently Long-form documents Long-form documents Struggles with scaling Struggles with scaling Handles efficiently Handles efficiently Real-time interaction

Often too slow

Much better latency Real-time interaction Real-time interaction Often too slow Often too slow Much better latency Much better latency Limited compute

Requires significant hardware

Runs efficiently at scale Limited compute Limited compute Requires significant hardware Requires significant hardware Runs efficiently at scale Runs efficiently at scale Ecosystem maturity

Extremely mature

Catching up quickly Ecosystem maturity Ecosystem maturity Extremely mature Extremely mature Catching up quickly Catching up quickly Implementation Example: Document Processing Pipeline Here’s a simplified example to illustrate the implementation differences 1. Shared Setup: Extract Text from PDF import pdfplumber
from transformers import AutoTokenizer

def extract_text(path):
    with pdfplumber.open(path) as pdf:
        return "\n".join(
            page.extract_text() for page in pdf.pages if page.extract_text()
        )

pdf_text = extract_text("freight_invoice.pdf")
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
inputs = tokenizer(pdf_text, return_tensors="pt", truncation=False) import pdfplumber
from transformers import AutoTokenizer

def extract_text(path):
    with pdfplumber.open(path) as pdf:
        return "\n".join(
            page.extract_text() for page in pdf.pages if page.extract_text()
        )

pdf_text = extract_text("freight_invoice.pdf")
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
inputs = tokenizer(pdf_text, return_tensors="pt", truncation=False) 2. Mamba-Based Inference from mamba_ssm.models import Mamba, MambaConfig
import torch

mamba_config = MambaConfig(d_model=512, n_layers=4)
mamba_model = Mamba(mamba_config)

with torch.no_grad():
    mamba_outputs = mamba_model(inputs["input_ids"]) from mamba_ssm.models import Mamba, MambaConfig
import torch

mamba_config = MambaConfig(d_model=512, n_layers=4)
mamba_model = Mamba(mamba_config)

with torch.no_grad():
    mamba_outputs = mamba_model(inputs["input_ids"]) 3. Transformer-Based Inference from transformers import BertModel, BertConfig

bert_config = BertConfig(hidden_size=512, num_hidden_layers=4)
bert_model = BertModel(bert_config)

with torch.no_grad():
    transformer_outputs = bert_model(inputs["input_ids"]) from transformers import BertModel, BertConfig

bert_config = BertConfig(hidden_size=512, num_hidden_layers=4)
bert_model = BertModel(bert_config)

with torch.no_grad():
    transformer_outputs = bert_model(inputs["input_ids"]) The Transformer setup works; but on long documents, you’ll quickly notice the difference in latency and memory usage. With Mamba, the same task runs faster and scales more gracefully, especially in high-throughput environments. Why This Matters in Practice Why This Matters in Practice What is happening here is far more than selecting a new model. It indicates something else: the engineering teams perceived to have ‘grown up’ in how they reason and design AI infrastructure. Rather than mindlessly optimizing benchmark scores, there is more product-thinking in the room: What can we accomplish in terms of design if inference is significantly quicker and cheaper?
What models could potentially enable us to ship with minimum impact to our infrastructure budget?
How efficient will this system be after six months, and after six product updates down the line? What can we accomplish in terms of design if inference is significantly quicker and cheaper? What models could potentially enable us to ship with minimum impact to our infrastructure budget? How efficient will this system be after six months, and after six product updates down the line? Not every team asking those questions will find SSMs compelling. But those that do, they absolutely will. What This Means for Product Teams Moving Forward What This Means for Product Teams Moving Forward Transformers won't be wiped out completely; SSMs don't have to. The right tool still depends on what you’re building. However, if the product handles lengthy documents, requires immediate feedback, and has to operate within moderate system requirements, SSMs should be considered. This change in approach is not as simple as moving from one model to another. The one switch that is needed is from "research driven decisions" to "product driven decisions". This change has been overdue for some time now. "research driven decisions" "product driven decisions" Amid the fast paced world of AI, these changes are not essential and desirable, for some of the most products that are beneficial and help you go live end up being overshadowed.

The AI Industry's Obsession With Transformers Might Finally Be Waning

Stories, News & Mentions about MINDMESH

Trending Topics

Classic

Neon Noir

Minty

Newspaper

HN StartUps