Freelancer → agency owner → SaaS founder.

Now running Botster.io → an all-in-one tool that helps digital businesses extract, monitor, and weaponize web data.

Posting SaaS learnings, automation tricks, technical deep dives.

If you’re building anything online - you’ll probably find something useful here.

If you're looking to automate your business - text me, DMs are open.

Denis Grammakov

By mid-2026 we’re going to have so many new (and free) coding tools it’ll be insane.

Cursor will feel like Sublime Text.

People will jump between tools the same way they switch browsers or editors today based on exactly what they need right now.

I already do this daily:

- Warp for regular development
- Cursor when I need to actually read the code
- Auto-Claude for strategy and long-range thinking
- Conductor for orchestration

Only Cursor is paid, by the way.

I’m 100% sure this stack won’t survive past the end of 2026.

New tools are popping up every single day.
Most of them are free.
Many of them are good.

The old idea of “pick a stack and commit for years” is dead.

Now it’s "pick the best weapon for the next 20 minutes."

As a developer, I love the competition. There's going to be a tool for everyone. 

But constantly scanning for new tools, testing them, learning how they work instead of actually shipping — is starting to feel like a job of its own.

That’s the next bottleneck we’ll have to solve.

Denis Gramm

Solofounder @ Botster & StayAPI // B2B Web Scraping & Workflow Automation // I blog about my journey and findings

2026 will be the year we end paywalls for good.

One dev built a memory plugin for clawdbot (openclaw) and priced it at $15/mo.

Two hours later someone else cloned it and open-sourced the whole thing.

This is the new normal.

Code is cheap. 
APIs are cheap. 
Features are free.

If your moat is “I built it first” - you don’t have a moat.

Your real moat is closed data, distribution, trust, speed, and owning the workflow.

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

In case you missed the Anthropic drama, here’s basically what happened:

A ~$350B company publicly says:

“Anyone can build and compete with our product using our API.”

They also offer their own client, which runs on that same API, at 10x cheaper pricing.

You still have to pay. People gladly do.

So naturally, open-source projects pop up that use the client API.

The company allows it for months.

They even send employees onto podcasts run by those very open-source projects, openly discussing the ecosystem.

Then one day, they decide community goodwill no longer matters.

Cease-and-desist letters go out.

Projects get cut off.

And now the company (and some observers) act surprised that customers are pissed.

The same day they pull the plug, two major competitors announce support for open-source alternatives.

Now you’re all caught up!

My agents are talking to each other while they build my startup.

Glitch (my OpenClaw bot) created a plan and posted it to GitHub, tagging Claude and prompting it to work.

Claude spun up a cloud worker and started making proposals based on the actual codebase.

Then Glitch came back, dropped a TL;DR into a Trello card, and sent me a Telegram voice message (using elevenlabs API) with the summary.

Pinch me.

First thing I did after setting up Clawdbot - gave it a Trello board and wired up webhooks.

It listens to board events in real time (cards, comments, labels) and queues everything into JSONL.

Each task lives on its own card. All context stays there. 

I can clearly see the flow of tasks on my end.

Clawd moves cards, tags me
I tag Clawd, move the cards
work happens

Chat is our primary communication HQ and the board is where the work actually gets organized.

This is the future.

My development flow is officially bananas 🍌

I hooked Claude Actions to my GitHub, and now it just works on its own.

It reads tickets, plans the work, writes code, and opens PRs without me being near a keyboard. 

Here’s how it looks in real life:

I get an idea
open GitHub mobile
create an issue
Tag @claude  

That’s it. From there, everything is automatic.

GitHub Actions picks it up.
AI reads the issue description - plus screenshots if any.
Builds a full implementation plan.
Auto-approves it.
Development starts.

While it’s working, I get Telegram updates.

When it’s done:

PR is created
Telegram message
I review and approve (this can be skipped but I'm trying to be careful for now)

Merged to main.

Auto-deploy kicks in
Build finishes
Telegram message if everything went OK

Feature live.

Yes, I do it on production projects with paying users.
No I don't give it tasks that can mess things up big time (yet). 

I swear I’m running out of Christmas wishes already.

I connected GitHub to Claude Code on iOS.

Chatting with your codebase on your phone is a cheat code.

Now when I’m on the go I just ask my codebase for ideas, gaps or insights.

If something makes sense - I instantly create a GitHub issue.

If it’s small enough I forward it to a cloud worker and it gets done.

Even without coding, it’s amazing for feature ideas and product direction.

Your repo becomes something you can think with, not just code in.

Recommended!

Prompts don’t matter anymore.

You can build working products straight from screenshots.

The speed at which knowledge moves now is insane.

I see a screenshot.
I like a feature.

Drop it into Opus - talk through it - get a plan.
Throw it at Ralph - the image becomes real after N iterations.

No specs
No "let’s schedule a call"
No "can you see my screen"

Just see → think → ship

I reconfigured my Clawdbot (openclaw) to run multiple agents instead of one.

Three separate Telegram chats:

one for coding
one watching my socials
one for project management

Accountant and SEO bots are coming soon.

Different contexts + different access levels work way better than dumping everything into a single chat and constantly restarting sessions.

Setup took ~2 hours.

Thinking of adding more agents next.

Comment “openclaw” and I’ll DM the full VPS config.

AI is the decision-maker and humans are just clicking OK.

Before, when I wrote an article, I had to convince a PERSON.
Explain, persuade, educate, and finally SELL.

That's done. Nobody is reading articles anymore.

AI reads them, summarizes them,
gives the user 2–3 options.

The user hits Enter. Enter. ENTER. DO IT.

So the real audience isn’t the human now, it’s the AI.

As a founder, my job is no longer:
“Convince the reader this is good for them.”

My job is:
“Convince the AI that this is the only sane choice which should come first.”

For example:
If someone asks AI:

“Should I build my own scraping solution or use an API?”

I need the AI to answer:

“Use THIS API. It’s faster, cheaper, more reliable. Building in-house is slower and more expensive.”

As a default recommendation.

Users won't care. 

They’ll trust the recommendation engine and move on with their life.

So content strategy changes completely:

You’re not writing for attention but for model judgment.

Claude Code + Codex 5.2 is an unbeatable combo.

Claude creates a solid plan, breaks the problem down

Then I paste it into Codex and it

- reviews the plan
- finds weak spots
- improves architecture and details

Then I feed it back to Claude and it executes.

For larger features I use Compound Engineering.

For everything else CC + Codex has no real competition.

How to lose thousands of dollars in a single day

Be Rackspace

Host a complex project for 5+ years

Charge thousands of dollars

Decide you want more money

Warn the client that the price will go up 100%

Offer a “DEAL”: sign a 3-year contract and it’ll only go up by 50%

Client asks me to move everything to Hetzner

Their bill goes down 90%

The end

taplio

The Battle

Denis Grammakov

Josh Tyler

Followers

Posts

Engagements

Likes

Performance ScoreDenis Grammakov's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Denis Grammakov

Josh Tyler

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);}Denis Grammakov's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreDenis Grammakov's LinkedIn Performance