I build open-source tools that turn domain knowledge into repeatable guidance for AI assistants.
West Greece, Greece
Joined June 2024
- Tweets2K
- Following662
- Followers519
- Likes1.2K
Your marketing dashboard can show every channel contributing to a conversion.
Search gets credit. Social gets credit. Email gets credit. Paid gets credit.
Who explains how the same customer became four success stories?
2
12
A thousand founders saying they love your product idea tells you very little about whether anyone will buy it.
Founders are unusually interested in products. That doesn't make them representative customers.
A lot of startup validation is founders validating each other.
1
7
Everyone is downloading AI agent skills. Some people seem to collect them like browser extensions.
How many have you actually read before giving them access to your tools?
A skill is executable trust. Treat it accordingly.
2
1
16
My precious AI skill, Second Read, has reached 107 downloads on clawhub @openclaw. What is your excuse for not giving it a try as well as a star to its open repo? github.com/C-Kapsalis/second…
1
2
75
I believe in building tools that find users where their decisions are made and help them organically instead of building new destinations. My pinned repos carry such ai skills and plugins you can check out github.com/C-Kapsalis
1
12
Calling builders and product people: Second Read Studio vs GoodPostal is live. Compare them and add a thoughtful verdict.
@FounderBattle
founderbattle.com/battle/sec…
2
31
⭐ Second Read just earned "Top 25 This Week" on @launch_llama!
#20 overall
tools.launchllama.co/product…
1
9
They say following on this platform does not matter when noone engages with your posts. But how do people with 10k+ followers cope with the fact that the responses to their posts get hidden under "Show probable spam"?
1
1
16
When I build an agent skill, I want it to encode the checks I'd otherwise do by hand. Then I can inspect the procedure, challenge the result and make the final call. What checks have you turned into reusable skills?
1
1
9
An agent skill that handles documentation needs a stop condition. What happens when the source is missing, the product behavior is unclear, or the task falls outside its instructions? If it keeps writing anyway, I wouldn't trust it with the docs.
1
7
“The campaign worked” tells me almost nothing. Show me the holdout, conversion definition, revenue source and measurement window. Without those, I can't tell whether the campaign created incremental sales or whether the dashboard just gave it credit.
1
15
An LLM can draft an SVG for a Power BI bullet chart. You can paste it into a matrix and get something that looks right. Have you checked that the displayed numbers still match the measure? That's where I'd spend the review time.
1
18
I use agents to search repositories and assemble a first draft. Then I move sections, add examples from the work and cut what the reader doesn't need. The first draft is the cheap part. Editorial judgment still takes subject knowledge.
1
12
Before installing a community agent skill, read its SKILL.md. What tools does it call? When should it run? When should it stop? The Scrapling team's official skill is a useful reminder to check the source instead of trusting a popular copy.
3
50
“The competitor charges €49” is a reason someone picked a number, not evidence that buyers will pay it. My value-pricing workflow treats €49 as a hypothesis: scan the market, test the feature premium, simulate buyers, then validate with real ones.
1
8
PBIP and TMDL make Power BI projects easier to version. They won't catch a broken lineage tag or a visual with no bound field. I'd put those checks in a pre-deployment lint. Finding them in production is an expensive way to test.
1
2
25
A 4.8x ROAS can look convincing and still be a bad reason to move budget. Check the conversion definition, attribution window and grain. Then ask the awkward question: did the campaign create incremental sales, or just claim credit?
1
1
12