About
I sell software online. Whether its consulting, enterprise sales, or software as a…
Services
Activity
758 followers
- Steve Hanov shared thisAnand is one of the smartest people I've worked with. He was implementing deep learning demos when we were all still using SVM.Steve Hanov shared thisThrilled to share that my co-author Anand Oka and I have published our book: Agentic AI Unleashed: A guide to designing, building, and deploying autonomous AI systems We wrote this book with a simple goal: help anyone — students, engineers, business leaders — build a clear, end-to-end understanding of Agentic AI. From foundational concepts to real-world deployment, the book walks through the full journey. We hope it finds its way into classrooms too — there is already some early interest in India that we are excited about. We had a wonderful moment launching the book at our Truveta Annual Homecoming — where we distributed copies to all Truveta employees and signed as many as we could. Means a lot to have the support of Terry Myerson and Truveta for supporting this launch and putting a copy in every employee's hands. One thing we feel strongly about: All author proceeds go to non-profits advancing education and access to opportunity. Every copy you read or share directly supports that mission. So please — if this topic resonates with you, pick up a copy, share this post, and help spread the word. That's all we ask. Deep thanks to BPB Publications and every reviewer, editor, and colleague who made this possible — including my wonderful co-author on this journey. If you're curious about where AI is headed and how to build for it, this book is for you. 🙏 #AgenticAI #AI #ArtificialIntelligence #Education #BookLaunch #Truveta #BPB
- Steve Hanov reposted thisSteve Hanov reposted thisThere are more important things happening now -- in fact, nearly everything is more important -- but you might enjoy my new word game CADGY. It's getting great response from players. It's different from the rest because it will actually improve your vocabulary, at least when it comes to 5-letter words. https://onelook.com/cadgy/
- Steve Hanov shared thisI thought I finally knew exactly who my SaaS was for. Then my users told me I was completely wrong. Last month, I pitched Eh-Trade (dot ca) at StartupCamp. I had the perfect, polished elevator pitch: "A tool that tells retail traders what stocks are going up, and exactly when their holdings need attention." The crowd loved it. I felt invincible. Then I sat down and actually interviewed three more potential users. Two of them were ex-finance professionals. They loved the back-testing and complex quantitative algorithms. If I could just save them from writing custom Python scripts, they would buy it instantly. The third user was just looking to graduate from Wealthsimple. His feedback? "This is terrifying. There are too many numbers. Just tell me what's good to buy!" As a founder, this is the ultimate trap. → The pros want MORE complexity. → The beginners want LESS. → And telling people what to buy is illegal. My wife actually solved the problem in minutes: Tiered UX. We need an interface that grows with the user. Show the simple health scores first. If they want to dig into the Black-Scholes options math, they can upgrade and unlock it. While I digest this identity crisis, my Waterloo intern Jimmy has been relentlessly shipping code. This week we launched: ✅ A Leverage Calculator to balance buying power and risk. ✅ An Earnings Data visualizer that already beats the major incumbent sites. ✅ Programmatic SEO so our backend data actually ranks on Google and LLMs. But I’ll be honest. Right now, it feels less like a unified product and more like a Frankenstein monster of cool, disconnected features. Creating something from nothing is incredibly messy.
- Steve Hanov shared thisMost people look for ways to extract value from their social network, mining their LinkedIn connections for hidden nuggets of gold. Successful entrepreneurs turn this on its head. It's all about finding the value that YOU have to GIVE to others in your community. Build social capital, and more opportunities that you can deal with will just show up. This is the real secret strength of the Waterloo ecosystem. That's why I made the PeepDeck app (peepdeck dot app on the web). I added a feature with weekly goals to reconnect with "weak ties". Research shows your most valuable connections come from the periphery. It's the people you barely know that have the most diversity and knowledge different from your inner circle. Thanks Sarah-Beth Bianchi -- your encouragement goes a long way. Networking is different for women in tech and also neurodiverse individuals -- and Sarah-Beth an absolute expert on both! She'll be giving a talk on April 16 in Waterloo.
- Steve Hanov shared thisAt the last Founder's dinner, run by Jesse Rodgers, the guest was Hongwei Liu, the founder of MappedIn. Someone asked him how he would start another venture if he had to today. "Like totally from scratch? Without any social capital?" He made a face I will never forget. It becomes clear very quickly that building anything in life is vastly easier when you have social capital. But sociological research shows that blindly clicking "Connect" on LinkedIn won't get you there. It just builds a database, not a network. Think about who you wish you had on speed dial for help and advice. It's the people who have a reputation for helping. They build communities. They promote the successes of others. They’re the ones volunteering and running the room. Here is the engineering problem: Human cognitive limits guarantee that your network will decay if you don't actively maintain it. So how do you become a super-connector without making it a full-time job? Map your layers: Research shows networks are structured in tiers (an inner circle of ~15, and a broader layer of ~150). Treat them differently. Track context: Keep track of what people actually care about, not just their job titles. Fight decay: Consistently follow up and add value to their lives long before you ever need to ask for a favor. Look for diversity: Research shows the people on your periphery should be from a diverse set of fields. They aren't reading the exact same things you are every day, and they are in industries with problems you haven't even heard about. That's why I built PeepDeck as a quick side project. It's a direct pipeline for building your social capital. It helps you map your layers, network mindfully, and prevents your most important ties from decaying over time. Successful founders all create their own luck. The more social capital you can bank, the luckier your projects will get. Give it a try. (Link in comment)
- Steve Hanov shared thisWaterloo birthed BlackBerry, Faire, and Tailscale. If you are in town for the Socratica Symposium hunting for the next billion-dollar startup, do not waste your time taking meetings in generic hotel lobbies. I tracked the top funding rounds out of Waterloo this year, and a massive percentage of them were quietly brokered over flat whites at these exact coffee shops. Here is the definitive guide to the best cafes to vet founders, dodge rival investors, and deploy your $500k seed round. Williams Fresh Cafe (University Plaza) Steps from the University of Waterloo campus, this is absolute ground zero for early-stage funding. Every single student and advisor knows exactly where it is. If they show up carrying a hardware prototype in a shoebox, you know you are in the right place. Just be prepared when they look you dead in the eye and demand a $25M post-money valuation on an uncapped SAFE. You will have to gently remind them that a Notion page is not a product. Balzac's Coffee Roasters (The Tannery) This one is painfully obvious, but mathematically necessary. Balzac’s is located at 151 Charles St W, directly inside the Communitech Hub in Kitchener. The founders you meet here are already savvy and highly accelerated. They will try to casually name-drop their latest YC interview while aggressively refusing standard board observer rights. You will literally bump into engineers from Google and D2L while waiting in line, so make sure you keep your term sheets hidden in your Patagonia vest. A Matter of Taste (Downtown Kitchener) Located near Kitchener City Hall, this spot puts you right near the local builder community. It is steps from the Builder’s Club, a central hub of the local developer scene. Order the Cafe Dulce. It is a brilliant, high-octane blend of condensed milk and a heavy shot of espresso. It provides the exact sugar rush you need while dissecting a founder's user acquisition costs. This is where you get to explain that pivoting from a Web3 crypto exchange to an "AI-driven workflow agent" does not automatically justify their burn rate. Push for your 20 percent equity target before their caffeine crash hits. Covenant Cafe (Erb Street) If you are discussing highly sensitive IP or finalizing a Series A, you cannot risk being overheard by competing VCs in Uptown. Covenant Cafe is far enough away from the university bubble to offer true privacy. They serve incredible fresh-baked treats and boast a quiet, charming atmosphere. It is the stealth-mode coffee shop for stealth-mode startups. Expect them to slide a printed mutual NDA across the table before they even order a blueberry scone. This is where you fund the deeply technical duo who refuse to update their LinkedIn profiles until Series B.
- Steve Hanov shared thisYour brokerage account is probably lying to you. If you bought AMZN five years ago and it dropped 20% last month, your portfolio still shows a comforting, bright green "+40% All Time." Brokers are designed to make you feel good so you don't close your account. But as a trader, that green number is a trap. It hides the fact that your capital is bleeding today. Welcome to Month 3 of building Eh-Trade.ca, a stock market tool designed to actually protect retail investors from themselves. Between running my other SaaS businesses, I fell behind on talking to users this week. But my Waterloo intern, Jimmy Zheng, has been shipping relentlessly. Here is what we just pushed to production: The "Bleed" Catcher: We turned our standard watchlist into a full portfolio tracker. It knows your positions, but instead of just showing you a sea of green, it actively sorts the stocks that need your immediate attention (the ones losing momentum) right to the top. Themes: Wall Street organizes stocks into 11 rigid "Sectors." Nobody thinks like that. You think in terms of "AI," "Crypto," or "Cybersecurity." We set an autonomous LLM running on a server in my basement to analyze and tag 19,000 securities into modern themes. Now you can search and compare companies that are doing similar things. And my basement is cozy warm. The Leverage Calculator: For our options traders, Jimmy built a tool that instantly compares up to four stock replacement strategies, balancing buying power against risk without doing manual Black-Scholes math. I pitched Eh-Trade at StartupCamp (huge thanks to Jesse Rodgers for running an incredible event), and the reception blew me away. It completely renewed my confidence in the core thesis: We have a product people want. The algorithm works (I’m up 60% personally). The problem now? Trust and Traffic. I am back to looking for brutal, honest feedback. If you know anyone who actively trades stocks or options, please tag them below or send me a DM. A 30-minute chat saves us 3 months of writing the wrong code.
- Steve Hanov shared thisI let open-source copycats like Mermaid.js and PlantUML get ahead of me. That stops today. WebSequenceDiagrams has a layout engine that can handle truly massive architectures with thousands of connections. To prove it, last week I added direct .HAR file support -- a direct ask from master of ad-tech James Strang. Now, you can directly open a log of a web page loading, and the engine automatically groups items into visual categories, letting you see the bottlenecks right away. But a great engine isn't enough if it's not where developers actually work. After talking to Cole Van Vlack at the builder's club to see exactly how he uses Mermaid, I realized something: The modern developer's workflow has walled itself off. If your tool doesn't live natively inside their IDE or their markdown notes, you basically don't exist. It doesn't matter that my proprietary layout engine can untangle massive enterprise architectures that make other open-source tools choke. If an engineer has to context-switch out of their editor to use my site, they simply won't do it. So, I built a VS Code extension that afternoon. It features syntax highlighting, adds support for markdown blocks, and puts a real-time preview right in the user interface as you edit. (You can install it right from the WebSequenceDiagrams landing page). Additionally, I submitted pull requests to the popular Markdown Preview Enhanced extension, and its author is officially making WSD syntax a first-class citizen in that tool. As a founder, my new goal is aggressive: make one integration every single day, targeting the exact tools where people are currently using Mermaid. Next up is Obsidian. Meanwhile, I'm feature-matching the existing tools. First up, I'm working on adding flowcharts using my own unique slant. But I need your help. If you currently use Mermaid flowcharts, or its other diagram types, I want to talk with you. Even a 10-minute conversation with a user can save me months of wasted work. Please DM me.
- Steve Hanov shared thisAI almost killed my 19-year-old SaaS business. Here is how I'm saving it using old-school code. WebSequenceDiagrams has been paying my bills since 2007. It quietly runs on a $5 Linode, doing one thing: converting text into beautiful UML diagrams. Then the 2026 "SaaSacre" hit. MRR dropped 14% year-over-year. Active users vanished. The culprits? First, LLMs. Modern AI assistants default to spitting out free Mermaid.js syntax for diagrams. Second, corporate "Zero-Trust." InfoSec teams no longer want their employees sending proprietary architecture blueprints across the internet to my random web server. The naive solution is to add a whiz-bang "Chat with your diagram" button. The pragmatic solution? Over-optimize. I looked at the data. My core asset wasn't the website. It was the proprietary C++ algorithmic layout engine I wrote years ago. While free tools like Mermaid use basic directed graphs that turn complex enterprise architecture into overlapping spaghetti, my engine actually understands visual logic. It's magic. Here is the 3-step technical pivot to outmaneuver the AI wave: 1. I'm reviving the "Enterprise server". I had switched to team subscriptions, but the winds have shifted once again. InfoSec is happy because the data never leaves their company network, and I don't need to shell out for Soc II compliance. 2. The Model Context Protocol (MCP) Instead of fighting AI, I became its backend. I deployed an MCP server. Now, when a developer asks Claude or Copilot to design a payment flow, the AI natively hooks into my superior layout engine to generate perfect, error-free syntax. 3. I'm leaning into its performance efficiency. It can import and make sense of a HAR file with hundreds of requests, letting you find bottlenecks in enterprise flows in milliseconds. The AI software commoditization wave isn't the end of indie hacking. It just means the era of lazy cloud APIs is over. If you want to survive, STOP bolting AI onto your front-end. Take your deepest, weirdest technical moat, shrink it down, and run it locally. Is the SaaS industry in trouble? What do you think?
Steve Hanov commented on a post 11h
It hit the right level. I enjoyed the refresher, without being overwhelmed with math. Colby described each concept in three abstractions: basic (intuitive understanding), engineering challenges (for an implementor), and research challenges (for a researcher) and it helped. Ti Guo always keeps things strictly on time.
- Steve Hanov liked thisI am deeply grateful to Jay Nanduri for the opportunity to work on our book "Agentic AI Unleashed" (https://a.co/d/08bN7yiS). It was a labor of love, pursued alongside the demands of our busy day-to-day professional responsibilities. We were driven by a desire to share our learnings about the fast-evolving and revolutionary technology of agentic AI — a true game changer not just for software engineers, but for all types of product development. We hope our contribution gives a new generation of innovators and entrepreneurs a meaningful head start. By directing all author proceeds to non-profit organizations advancing education and access to opportunity, we also hope our work has a compounding effect beyond these pages. We are grateful to all our colleagues and friends who supported us throughout this endeavor. A special thanks to BPB for their continuous and dedicated support as we authored and published this work. We hope you enjoy the book — and please spread the word! #BPB #Truveta #AgenticAIUnleashed #AI #BookLaunch #EducationSteve Hanov liked thisThrilled to share that my co-author Anand Oka and I have published our book: Agentic AI Unleashed: A guide to designing, building, and deploying autonomous AI systems We wrote this book with a simple goal: help anyone — students, engineers, business leaders — build a clear, end-to-end understanding of Agentic AI. From foundational concepts to real-world deployment, the book walks through the full journey. We hope it finds its way into classrooms too — there is already some early interest in India that we are excited about. We had a wonderful moment launching the book at our Truveta Annual Homecoming — where we distributed copies to all Truveta employees and signed as many as we could. Means a lot to have the support of Terry Myerson and Truveta for supporting this launch and putting a copy in every employee's hands. One thing we feel strongly about: All author proceeds go to non-profits advancing education and access to opportunity. Every copy you read or share directly supports that mission. So please — if this topic resonates with you, pick up a copy, share this post, and help spread the word. That's all we ask. Deep thanks to BPB Publications and every reviewer, editor, and colleague who made this possible — including my wonderful co-author on this journey. If you're curious about where AI is headed and how to build for it, this book is for you. 🙏 #AgenticAI #AI #ArtificialIntelligence #Education #BookLaunch #Truveta #BPB
- Steve Hanov liked thisSteve Hanov liked thisStarting a 1hour/wk x 6wk, Hands-On "Reinforcement Learning 0 to Hero" study group, from basics to RLHF, World Model, and MiniMax & Others‘ RL-for-agents Framework, and coding up RL models, with our community friend teacher Colby (Ziyu) Wang (RL class TA in TMU) and 100+ registered classmates. Open to developer, ML engineer, and ML researchers. Reinforcement learning matters because it helps AI agents make decisions in robotics, games, healthcare, and recommendation systems, and could be an important building block on the path toward AGI. Also sharing the overwhelming feedbacks received during the session: “This presentation was amazing, magnificent.” — Miris “Thank you for a very nice presentation.” — Sachin “Great presentation.” — Blush Just finished 1st session last wk. Feel free to watch the recording: https://lnkd.in/etCkCThs And watch the future recording in our community “Paper Reading Group” Section: https://lnkd.in/enYhHk_4 Register for next wks class at: https://lnkd.in/gpGQExxe Code repo for the class: https://lnkd.in/eERpA8ju Here is the Agenda: Week 1 — Apr 4: Introduction to reinforcement learning — how AI learns through trial and error. Week 2 — Apr 11: Value + Policy based RL — how AI scores actions and learns to choose better actions. Week 3 — Apr 18: Value & Policy Hybrid RL — how AI combines scoring and action-taking. Week 4 — Apr 25: RL in Agent — how MiniMax Forge trains agents for complex tasks. Week 5 — May 2: RLHF — how human feedback helps AI become more helpful. Week 6 — May 9: World Model RL — how AI imagines outcomes before taking action. Thank you to our course co-creators Colby (Ziyu) Wang, Alp Güneysel, Will Huang And classmates: Ruthvika Dasaraju, Karthik M.,Aahan Rashid, Sachin Ganpule, Madhu P., Shayan Shafquat, Vinaya Sangeeta Lahari Baswa, Krystal Maughan, Alp Güneysel , Deepto Dutta, Devica Verma, Iris Guo 🌎 Armaan Amatya, @Miras, Blaž Blokar, Brenda Li, Tristan Udby, Dwayne Khallique Jones, Divya M., Steve Hanov, sachinganpati ganpati, Simeng HU(Chloe), Vinaya Sangeeta Lahari Baswa, Aditya Bharadwaj, YunQi Liu #ReinforcementLearning #RL #RLHF #WorldModels #AIAgents #MachineLearning #ArtificialIntelligence #AGI #LLM #AIEducation #AICommunity #MLResearch #MLEngineer #Developers #StudyGroup #PaperReading #MiniMax #AgenticAI
- Steve Hanov liked thisMost people look for ways to extract value from their social network, mining their LinkedIn connections for hidden nuggets of gold. Successful entrepreneurs turn this on its head. It's all about finding the value that YOU have to GIVE to others in your community. Build social capital, and more opportunities that you can deal with will just show up. This is the real secret strength of the Waterloo ecosystem. That's why I made the PeepDeck app (peepdeck dot app on the web). I added a feature with weekly goals to reconnect with "weak ties". Research shows your most valuable connections come from the periphery. It's the people you barely know that have the most diversity and knowledge different from your inner circle. Thanks Sarah-Beth Bianchi -- your encouragement goes a long way. Networking is different for women in tech and also neurodiverse individuals -- and Sarah-Beth an absolute expert on both! She'll be giving a talk on April 16 in Waterloo.
- Steve Hanov liked thisSteve Hanov liked thisProposed 5X Cheaper Agentic Coding Pipeline, inspired by leaked Claude Code. See Code Demo below Just had a great group discussion with friends on lessons learned from Claude Code leak. One of the biggest takeaways is this: “Agent memory” and multi-coding-agent workflows may be much simpler than we thought. Our full session recording: https://lnkd.in/eqQWPTBZ Our code demo on building a simple Cognition's Devin AI clone with LangChain LangGraph: - SpecKit version code demo: https://lnkd.in/ey7sphkC - OpenSpec version code demo: https://lnkd.in/eGU-aqZ5 From our group discussion, we came up with a structure inspired by a few ideas from Claude Code: - memory as Markdown - multi-agent coordination through shared tasks - a shared artifact layer between planning and execution That led us to propose this workflow: - Claude Code for planning and architecture - OpenSpec (YC W26) / GitHub SpecKit as the shared memory layer between tools, for sharing specs as markdown - MiniMax M2.7 (via OpenCode) for faster, lower-cost execution SpecKit Workflow (Our project implemented in SpecKit: https://lnkd.in/ey7sphkC) - [In Claude Code]`/speckit.constitution` → static project rules - [In Claude Code]`/speckit.specify` → overarching feature requirement - [In Claude Code / OpenCode]`/speckit.plan` → structured implementation plan - [In Claude Code / OpenCode]`/speckit.tasks` → task breakdown + dependencies - [In OpenCode]`/speckit.implement` → execution + iteration OpenSpec (YC W26) Workflow (Our project implemented in OpenSpec: https://lnkd.in/eGU-aqZ5) - [In Claude Code] `/opsx:propose` → define the change proposal [Claude Code] - [In Claude Code] `/opsx:explore` → inspect codebase / gather context - [In OpenCode] `/opsx:apply` → generate and apply implementation artifacts - [In OpenCode]`/opsx:archive` → finalize and archive completed change Architecture [ Claude Code ] (Planning / Architecture) | v [ OpenSpec / GitHub SpecKit ] (Shared memory as specs in Markdown + shared task layer) | v [ OpenCode + MiniMax ] (Execution / lower-cost coding) Would love to hear your thoughts and feedback — especially if you see a better way to structure the shared memory / task layer across agents. Thanks Divya Ann Kurien, Emeka Anekwe, Colby (Ziyu) Wang for helping with the code, Arnold Xu for eariest inspiration, and Jay Shah, Colin K., Armaan Amatya, Karthik M., @Paul Staadecker, @Tom Nguyen, Di Nguyen, @Ahmed Abdullah, Shake D., Alp Güneysel, Javier Cortavitarte, Luisa P., and Osledy Bazó-bazo and other friends for attending the workshop! #ClaudeCode #AICoding #CodingAgents #AgenticCoding #OpenSourceAI #DeveloperTools #LLM #AIEngineering #SpecDrivenDevelopment #OpenCode #MiniMax
Experience & Education
-
Hanov Solutions Inc.
View Steve’s full experience
See their title, tenure and more.
or
Patents
-
Apparatus, and associated method for facilitating communication connection between a mobile station and a communication network
Issued US 8218478
See patentIn the early stages of UMA/GAN, some equipment used version 3 of the standard and some used version 4. The ones using version 3 wouldn't accept version 4 messages.
The patented solution is to try one first, and if that doesn't work, then try the other one. -
Apparatus, and assicated method, for selecting and negotiating frame size of communication data communicated in a radio communication system
Issued US US 2008/0310448 A1
See patentWhile implementing the UMA/GAN technology I noticed that data transfer was very slow. Basically the IP packet size submitted to the protocol stack was a few bytes too long causing segmentation and wasted ethernet frames. When GAN is in use, I found a way to tell the network to use the optimal frame size for packets. Nobody uses data transfer over GAN anyway -- if you have a wifi signal, its better to just use that!
-
Apparatus, and associated Method for Lengthening Data Communicated in a Radio Communication System with Padding Bytes
Issued US US 2008/0205648 A1
See patentAn unfortunate title that was chosen by the lawyer... A pedantic bug was reported in BlackBerry from a carrier test lab. After investigating, I found it was due to differing interpretations of the 3GPP standards for 2G EDGE technology. Doing it their way would have made performance slightly worse. I filed a standards change and associated patent so that my implementation would be the correct one.
-
Method and system for detecting data congestion and applying a cell reselection offset
US 7747256
See patentThis is one is practical and really improves things when using 2G on blackberry. I fought hard to get the lawyers to understand it.
The basic idea is that when there are a lot of people using smartphones in a given area (eg, the RIM campus or a stadium) then data transfer will be slow. In that situation using a weaker cell makes sense and can give you the advantage.
Of course it makes no difference if everyone else is using blackberries too! -
Method for receiving and managing a downlink radio link control data block
US 8213375
See patentDuring data communication, sometimes a bit would be flipped over the air and the block would be corrupted but still pass CRC. Sometimes the bit is the one that ends the transfer and that's bad. I just added some code to check that all the bits made semantic sense and throw the block away if not. Also I patented this 1 line IF statement.
Projects
-
Zwibbler.com
- Present
See projectLet users draw on your web site with this customizable drop-in solution. I have been licensing the code and customizing it for use in medical annotations, scientific notebooks, educational apps, and T-shirt designs.
Among the first demos of the HTML 5 canvas as a graphics editor, this web site was also used as a test by the Internet Explorer 9 team. -
RhymeBrain.com
- Present
See projectRhymebrain.com is the world's largest rhyming dictionary and has 100,000 weekly visitors. It has 2.6 million words, and works in eight languages. I use machine learning to derive the pronunciation of unknown words, and a proprietary algorithm to calculate how much two words rhyme. Revenue comes from ads.
-
WebSequenceDiagrams.com
- Present
See projectWebsequencediagrams.com lets you create sequence diagrams in seconds instead of minutes. Join the hundreds of active subscribers to get premium features and support.
-
Various Windows Apps
See projectAn app has to have a single purpose, and be easy to use. I used this to achieve massive adoption of my freeware applications for Windows, which were profiled in PC Magazine and radio talk shows. Successes include Banshee Screamer Alarm and PhotoWipe.
Languages
-
Mandarin
Limited working proficiency
-
English
Native or bilingual proficiency
Recommendations received
1 person has recommended Steve
Join now to viewExplore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content