Greetings followers of ClawCognition and OpenClaw,

I want to provide a transparent Product Hunt update.

We just released a production ready RAG project-template for Bult.ai.

If you want to deploy a serious Retrieval Augmented Generation system, not a toy demo, this is for you.

With the advent of clawdbots, it's as if we've all lost our inhibitions and "put our lives completely in their hands."

I'm all for delegating work, but not giving them too much personal/sensitive stuff to handle.

We re making three frontier open-source models free* on Zo until the end of February. GLM-5, Kimi K2.5, and MiniMax M2.5 are now free to use on Zo. And we ve increased the AI usage limits on our free plan significantly.

We re incredibly excited about recent progress in open-source models. Three labs dropped big releases just weeks apart, ahead of the highly anticipated DeepSeek R2. These open models are quickly catching up to closed models like ChatGPT and Claude, which are much more expensive.

As usual, Y Combinator came up with segments that are worth investing:

1. Cursor for Product Managers
2. AI-Native Hedge Funds
3. AI-Native Agencies
4. Stablecoin Financial Services
5. AI for Government
6. Modern Metal Mills
7. AI Guidance for Physical Work 8. Large Spatial Models 9. Infra for Government Fraud Hunters 10. Make LLMs Easy to Train

Since I haven't been able to meet my work goals very well in the last few quarters, I now plan to approach them more systematically and not push myself too hard on work goals, as that ultimately led to problems that made my plan less sustainable.

So here is my structure and list:

New AI models pop up every week. Some developer tools like @Cursor, @Zed, and @Kilo Code let you choose between different models, while more opinionated products like @Amp and @Tonkotsu default to 1 model.

Curious what the community recommends for coding tasks? Any preferences?

Source.

Once upon a time, developing for Apple was an exciting, rewarding challenge. But lately, that relationship has soured.

Apple has transformed into a trillion-dollar giant that sees developers not as partners, but as a resource to control, extract from, and when convenient ignore.

This post is actually inspired by a tweet from @sandradjajic + an update here on PH from @chrismessina.

A few days ago I saw this:

Last year we hired a design agency to build our marketing site for @Basedash. They did an incredible job. The headline makes it sound like I'm dunking on them, but I'm not. The site was genuinely great. They built it in Framer so we could manage content ourselves, which was a completely reasonable bet at the time (and something we explicitly asked for).

Hot take on AI: The best AI tools will be invisible.

I used to think the best tools were the ones that had lots of features.

But making a tool disappear is more powerful than building endless features.

After our first launch on Product Hunt, our team spent a little over a month upgrading the product. There were major changes to the UI and several new features added, so the process took time from discussions and redesigning the interface to testing, fixing bugs, and updating AI prompts.

We re also a very small team, so everyone had to push themselves to give 200%. Time and resources are limited, and at the same time, we also had to work on securing funding for the next six months to keep the team running and continue developing the app.

First of all, I want to thank you for voting for us in yesterday's launch - you still can, the week is not over. HERE

Second thing, I summarised some things that I realised, reflected on, and maybe should have known sooner:

Hey everyone,

I've built my product around traditional SaaS pricing (monthly tiers), but I m starting to wonder if that model is getting outdated, especially with more AI-powered and compute-heavy tools entering the market.
That shift requires real architectural changes, instrumentation, metering, billing logic, and UI changes, not just pricing tweaks. It s something I m starting to seriously think about for my own product.

In particular, AI usage has real COGs (every prompt costs money), and I m seeing more platforms experimenting with usage-based models, or hybrids like SaaS base + usage + overage.

For those of you building AI or compute-intensive tools:

Linear just launched a new linear.app - a new source of inspiration? WDYT?

We re officially in nominees season for the Orbit Awards: AI Workflow Automation.

This is something I ll find out in just a short while, one week from now (Jan 28), as I m about to re-launch a digital detox app. If you want, follow, maybe you will be on watch of my steps and activities

However, that s not the main point of this post.

Today, I came across an article on TechCrunch: The great computer science exodus (and where students are going instead).

It shows that UC campuses saw a drop in computer science enrollment for the first time since the dot-com crash (6% in 2025, 3% in 2024), but students are shifting to AI-focused programs.

We've surpassed the singularity.

Hey PH

We're launching Rover on Feb 25th, but the preview is live right now and we want your honest takes before we go big.

What it is: One script tag on your website your users get an AI agent that takes real actions inside your UI. Clicks buttons, fills forms, runs checkout, guides onboarding. Through conversation.

<script src="https://rover.rtrvr.ai/embed.js"></script>

Today I received a question about proper launch preparation:

Nika, how long should this take us?

And I didn t have a clear answer.

When people started upgrading to PRO

I should ve felt great.

Instead, I felt Supaguard still wasn t feature-packed enough.

The question for 2026 isn t whether enterprises will use AI in code review.

They will.

The real question is:
what makes an AI code review tool enterprise-ready?

I just fixed and rebuilt an entire AI automation flow without opening the flow editor once.
Here's what happened:
We have a YouTube script generation pipeline on Serenities. It takes a keyword (or YouTube URLs), generates  an outline, then writes each section with precise word counts. Two branches one for URLs, one for direct topics.
The URL branch was broken. Not obviously broken. Subtly broken it was jumping straight from transcript cleansing to a single Claude call with no outline step, no section iteration, no word count targeting. Just a raw prompt that guessed at length every time.
I asked Claude to check it via MCP.
Within seconds it had read the entire flow every node, every edge, every config. It found:
- A broken template variable resolving to [undefined] instead of the word count
- Extended thinking enabled on a cleansing node (pointless, just burning tokens)
- The URL branch missing 7 nodes that the keyword branch had
- The MCP help guide serving 5 thin topics with almost no real documentation
Then still without me touching the UI it:
Added 7 new nodes to the URL branch (outline prompt, Claude outline generate, JSON cleaner, iterator,
section prompt, section writer, text aggregator)
Rewired all the edges correctly
Updated the webhook response to pull from the new pipeline
Rewrote all 5 MCP help topics with complete documentation utility node configs, template variable syntax,
Claude toolInputJson templates, debugging guides, everything
One conversation. Zero UI clicks. The flow went from broken to production-ready.
This is what MCP actually unlocks. Not just "ask AI to explain your code." Ask AI to read your live
infrastructure, reason about what's wrong, and fix it programmatically, precisely, at the API level.
We built MCP support into Serenities so AI agents can build and manage flows on behalf of users. Today I used
it on our own platform. It worked exactly as intended.
The future of automation isn't drag-and-drop. It's conversational.