I've spent nearly 25 years building software and leading product and engineering teams — from scrappy bootstrapped ventures to global public companies. 

Through successes, failures, and a lot of learning, I’ve been fortunate to work alongside world-class engineers, creatives, and leaders. That experience taught me what truly moves the progress needle: helping teams and people cut through the noise and align around shared context and meaningful work.

Today, I’m channeling that into Steady — an AI-native platform for coordinating digital teams without the overhead. Steady implements a set of open-source principles I helped develop called Continuous Coordination. Steady is a modern, human-first approach to work that aligns people with real-time context and helps teams deliver better and faster — without burnout.

Henry Poydar

Experienced Product Leader with 10+ years of experience driving growth and success for startups and enterprise businesses. Currently focused on creating AI-powered products and experiences that leverage multi-model (and modal) systems. A proven track record in both B2B and B2C SaaS environments, thriving at the intersection of technology, strategy, and innovation. 

Thriving in fast-paced environments, I relish the opportunity to tackle complex, challenging problems - the harder, the better. I excel at guiding cross-functional teams through the full product lifecycle, from conceptualization to launch and beyond, ensuring that data-driven insights and a relentless focus on the customer underpin every step.

My general product mindset is rooted in a deep belief that great products aren't just about what they can do; they're about doing meaningful things that meet people's needs in a simple, intuitive, and delightful way.

Mike Bal

Good job Microsoft. 

This is what losing with AI looks like. 

If you surface AI inside a product... it should be able to interact with the product on behalf of the user. 

If you surface AI without giving it the ability to make the user's life/task easier, you are telling them that you really only care about checking the "we have AI" box. 

I'm going to hold back my other feelings on Microsoft UX... for now.

So many people focus on what AI can automate, and I think that’s a very short sighted perspective. 

If you want to see impact you should focus on:

How can AI help improve the outcome?

How can AI help improve the quality? 

How can AI help expand the scope?

Sometimes the answer includes automating parts of the work, but 9/10 times I use it to take what I’m working on a few steps further. 

More research. 
More challenging perspectives. 
More data analysis. 
More polish. 
More delight.

Discovered this really clever onboarding trick from Neo4j

I thought it was a required sales form to get the free instanc up and running and maybe it was but...

This was different. 

When I typed my name, a node appeared on screen. 
Added my company, and suddenly there's a relationship connecting them. 
Kept filling out the form, and the graph kept building.

Show me > Tell Me

“Should we let users debug it themselves?”

Traditional thinking would say this is a terrible experience and that the user should never have to solve a problem themselves. 

We are in 2026 and more tools are adopting non deterministic flexibility in the product over rigid functionality, and it’s the right choice in many (not all) cases. 

This is t traditional debugging but after a few attempts at triggering research and skill use on the mobile app I realized it was just stuck. 

Instead of giving up, I simply called it out and it fixed the problem. 

IMO this turns a “bug” into an experience I understand and gives me more confidence in the product. 

I wouldn’t call it ”delight”. That would look more like Claude catching itself when the active thread stalled out, but that’s ok. Turning a poor experience into an empowering one is a major win. 

Kudos to you Anthropic team.

graphical user interface, text, application, chat or text message

Am I being overly critical?

Google DeepMind has put out some fantastic models and is leading in terms of developer experience - kudos to people on the team like Paige Bailey and Logan Kilpatrick 

The consumer facing Gemini app (web and mobile) from Google is not good and makes the models feel lower quality and less capable than they are. 

Very basic things like better thread context management, user specific memory, and shifting requests within a thread would go a long way. 

Things I hate:

Get an image and ask it to make subtle changes to the image in the same thread. You get back the exact same image with no changes but it thinks it did what you asked. 

Ask it to use the thread as context for something new e.g. a document, video, etc. misses half the thread. 

No way to branch context via project folders, thread groups, etc. so you have to give it all the context every time. 

They have done a good job of integrating across the Google ecosystem, so you can use drive for context management and work around a few of these things…  but that’s more work than it should be considering the experience across key competitors.

Appreciate Claire Vo  breaking down Clawdbot/Moltbot setup, risks, etc. and being 100% real

I have a separate dev computer now I typically use for testing new things. Most of the time I don’t hesitate to try new tools but this thing is as scary as it is exciting. 

I had a very similar experience during setup and decided to pump the breaks. 

IF I come back to this, I’ll clean an older computer, limit logins, set up a dedicated 1Password vault, etc. 

Being honest… as fun and exciting as it looks, I probably won’t come back to it for a while or at all. 

Instead, I’m doubling down on agents and skills. 

In terms of using AI effectively, it takes a bit of back and forth with prompting, context design, etc. to get it to do the thing you want the way you want it done. 

We are at a point where the big players are consumer-ising  agentic AI and most people (maybe even me) are sleeping on agents and skills. 

I don’t see a world where non techies/tinkerers are actually building/maintaining agents and skills, but I do see one where memory turns into awareness and the models can optimize the experience and costs by creating agents and skills where they see fit. 

Manus AI just took a big step this week by letting users turn a specific prompt/thread/outcome into a repeatable skill with a click or two. Not just the prompt but scripts, tool use, templates, etc. 
 

 https://lnkd.in/gVnz47e7

Open video: My honest experience with Clawdbot (now Moltbot): where it was great, where it sucked on youtu.be

My honest experience with Clawdbot (now Moltbot): where it was great, where it sucked

taplio

The Battle

Henry Poydar

Mike Bal

Followers

Posts

Engagements

Likes

Performance ScoreMike Bal's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Henry Poydar

Mike Bal

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Mike Bal's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreMike Bal's LinkedIn Performance