This is revolutionising how I work now.
Building my very own vibe coding & working tool.
I saw cnvs took the inspiration and wanted to make my own custom one, maybe i'll open source this?
I'm using this to run my company Trigify now. Going to keep adding to this daily as a passion evening project.
Celebrating small wins is hardwired into "Achieve," our goal-tracking sample app.
We wanted the simple act of checking off a task to feel like a celebration. Instead of the standard #Material ripple, we used Styles to drop in a custom, high-contrast completion animation—specifically, a brand-colored ring that expands and fades on tap, paired with a subtle pressed-state gradient.
It lets us focus entirely on polishing those custom, rewarding micro-interactions while Material handles the foundation.
#GoogleDesign #MaterialDesign #JetpackCompose #AndroidDev #UXDesign #InteractionDesign
I'm thinking about cancelling my Anthropic sub and going to Grok?
Fable & Opus have been so bad that i'm now only using Sol 5.6.
Feels like Grok 4.6 & Sol 5.6 could be my new drivers?
Hermes inside of Buzz is a great experience. I love the Hermes desktop app but this takes it to a new level allowing me to control other harnesses like claude code & codex etc.
Open Source version of Clay - https://t.co/LQY7Ry14kU
Been contributing a tone to this product as using it for Trigify.
Few cool neat features:
- Taps into your local subscriptions i/e Codex, Claude, Cursor etc
- Native pipeline feature within
- Has a cloud option for webhooks etc
AI has played with my mental health.
Let me explain why.
I use AI heavily through lots of different agents that I’m running in production.
My productivity has skyrocketed and it’s added immense value. BUT
The biggest challenge I have is this: what I’ve done is never enough.
Let me explain. I can now do 10 tasks at once and usually complete in a day what would have taken me a week.
Because of that, I’m left at the end of each day wishing I’d done more, feeling frustrated with myself for not pushing harder.
No one talks about this, but there’s a real high to being so efficient and seeing the outputs, and you’re left chasing that each day.
We thought AI would help with burnout, but I see it making things worse if you don’t catch yourself doing this.
@Teknium@PedjaDrazic@NousResearch@Teknium the other thing is support mutli profiles when using remote via the app.
I've added this into my local app now to support it but could be something to think about.
I pointed Claude Code at my biggest competitor's socials. 4 platforms, 165 posts, one knowledge file my agents now reuse everywhere.
Setup was simple. Told Claude Code to use the Trigify MCP and monitor Brandwatch across Substack, X, Reddit and YouTube. It created the searches automatically.
Then it analysed every post for:
→ Positioning and messaging themes
→ Product features and launches
→ Wins and traction signals
→ Frustrations and pain points
→ Pricing and value perception
→ Audience profile and sentiment
165 posts distilled into one markdown file.
But the file doesn't sit in a folder doing nothing. It becomes context for every other agent I run inside Claude Code.
Writing an email sequence? It pulls competitor context.
Building a landing page? It knows how they position.
Drafting a blog? It references what's resonating.
One knowledge file. Reused across copy, social, email, blog, strategy.
Stack five or six of these and your agents understand the whole competitive landscape before they write a single word.
Run the same prompt with the same model and the same skill. Wire up context + memory underneath and the output stops sounding generic
Prompt -> https://t.co/TjO8GuzRv3
Two layers underneath the model do all the work:
→ Context engine (internal + external)
→ Memory system (raw → belief → memory)
Skills sit on top. Most people build skills first, which is why their agents are mid.
External context = market signal.
X, LinkedIn, TikTok for high-velocity. Reddit, Substack, YouTube for depth. Pulled through https://t.co/HSiBb4mtxt.
Internal context = HubSpot, Notion, Slack, Granola, Fireflies, Featurebase.
The agent reads what I read.
The belief check is the part nobody builds.
A signal only becomes a memory if it appears 3+ times across raw data, or I state it as a hard fact.
One viral tweet ≠ a belief.
Without this filter, your agent is confidently wrong about everything that trended yesterday.
The LLM doesn't matter anymore.
Every founder I talk to is still picking sides. GPT vs Claude vs Gemini.
Pick the smartest model on earth and feed it nothing about your business and it still produces generic slop.
Here's what actually matters in 2026 →
Hermes agent has become my full time Ironman coach for my first race I have in August..
Its hooked up to Strava + Whoop.
It created me a 30 week plan which I've been executing daily on.
Every day its checks how I'm doing against my Strava & Whoop data and adjust the plan based on how well or badly I'm doing.
Here's its latest words of advice to me about completing it ahead of the cut off time.
I still feel like its going to be a survival march tbh..
The important part isn't the email. It's the timing.
48 meetings in 2 weeks from 120 leads because I reached out when they were warm.
More signal → context → action.
Phase 3: Claude builds the automation
Not a Zapier flow. The system watches the signals, qualifies the context, and only fires when something useful actually happens.
Phase 4: signal fires, workflow runs
→ enrich the person
→ find their email
→ research person + company
→ understand the exact signal
→ draft a personalised email
→ push to Instantly + Smartlead
→ notify the team in Slack