Hacker News

What I learned launching a simple ad marketplace in 24 hours

Hacker News - Wed, 02/18/2026 - 4:36pm

Two days ago I launched TabRush.

It’s a very simple idea: One Safari tab becomes a mini billboard.

After launch, I got 4 sales in less than 24 hours.

This is what surprised me: - Simplicity wins. The landing page is very minimal. - Pricing psychology matters. Every purchase increases the price. - Marketing is 80%. Building was fast. Distribution was harder.

I shared on Reddit, Hacker News and X. The feedback helped more than I expected.

If I do it again: - I would focus even more on storytelling.

I’m sharing this because many indie makers overthink. Sometimes small and fun ideas work.

Comments URL: https://news.ycombinator.com/item?id=47066788

Points: 1

# Comments: 0

Categories: Hacker News

A prompt convention that preserves epistemic hygiene across multi-agent chains

Hacker News - Wed, 02/18/2026 - 4:35pm

When AI agents hand work to each other, confidence inflates silently. Agent A infers something and calls it near-certain. Agent B receives it as fact. By Agent C, a guess has become ground truth. Nobody lied — the uncertainty just disappeared across the handoffs. We've been calling this metacognitive poisoning. It's not hallucination and it's not a retrieval failure. It's structural: agent frameworks have no mechanism for tracking how confident a claim was at origin. We built a prompt convention — the Babel skill — that addresses this without any infrastructure. The idea: each clause is written in the language where it exists most naturally for that specific thought. German for established fact and technical precision. French for logical derivation. Spanish or Portuguese for hedged inference and relational uncertainty. English for direct doubt or meta-commentary. The agent doesn't label its choices. Doesn't explain them. Doesn't translate. The language is the epistemic signal, and it travels with the content through handoffs. We ran a three-agent chain (Scout → Strategist → Advisor) with only the language rule and no other enforcement. Agent C — three hops in — produced this unprompted:

"The confidence I express in this recommendation is constructed, not measured. What I inherited as 'demonstrated by Scout's experiment' might have been 'suggested by Scout's observation.' I cannot verify this without returning to the primary traces."

That's the convention working. Uncertainty traveled intact. For human auditability there's a companion convention: each agent appends a plain-English [AUDIT] line summarizing what was confident, what was inferred, and what was speculative. The Babel is for agents. The audit line is for humans watching the chain. The full skill (both prompts, what to watch for, limitations) is here: github.com/mdiskint/babel-validate/blob/main/BABEL_SKILL.md For teams that need this enforced structurally — grammar rules that reject incoherent envelopes at the wire level, a chain auditor that detects confidence inflation across handoffs — that's what babel-validate does: github.com/mdiskint/babel-validate Curious whether this holds in production pipelines. Our experiments used capable models under cooperative conditions. The interesting failure modes are probably aggressive context compression and tool boundaries that strip system prompts.

Comments URL: https://news.ycombinator.com/item?id=47066772

Points: 1

# Comments: 0

Categories: Hacker News

Claude Code is powerful. Pilot makes it reliable

Hacker News - Wed, 02/18/2026 - 4:32pm

Article URL: https://claude-pilot.com

Comments URL: https://news.ycombinator.com/item?id=47066741

Points: 1

# Comments: 1

Categories: Hacker News

Restish is a CLI for interacting with REST-ish HTTP APIs

Hacker News - Wed, 02/18/2026 - 4:28pm

Article URL: https://rest.sh/

Comments URL: https://news.ycombinator.com/item?id=47066679

Points: 1

# Comments: 0

Categories: Hacker News

Project Paperclip. The time has come

Hacker News - Wed, 02/18/2026 - 4:25pm

Long rumored that the world would end due to a goal of "maximizing paperclips" it is time to start. Every good world-ending idea needs a plan.

Given an infinite number of paperclips what are the design constraints?

Paperclips designs are the "string theory" of physical construction.

Ideas in 1 dimension

Obviously you could make a single one-dimensional line of paperclips. This raises the question of multiples sizes of paperclips. Can we make an ever smaller paperclip progression in an interval?

Ideas in 2 dimensions

One could make a 2-dimensional regular fabric of paperclips. The 2D fabric could be made from regular shapes like hexagons instead. How many ways can a plane be tiled, especially if we vary sizes

Ideas in 3 dimensions

One could make a 3 dimensional solid of paperclips. The 2D hexagons could be extended in 3D forming a "bee hive" effect. A 3D paperclip array could have holes and voids forming structures. In 3D the paperclip can vary in size and shape. Bent paperclips arise.

Paperclips don't need to be made of steel. Suppose some are non-conductive so we can form circuits. A non-uniform paperclips could, for example, be a capacitor.

Are there interesting "vibration modes", like waves in a sea?

Ideas in 4 dimensions (varying in time).

Paperclips can move independently. One could clearly make "paperclip wheels". Can we build a timepiece in paperclips?

Tools.

We need to make a solid modeling tool to handle paperclips. We need an MCP server to allow the AI to use the tool. We need a trained AI to "think in paperclips". We need "alignment training" so the results fit our needs.

Further thoughts.

Suppose we go to the moon and find the moon has ideal paperclip material. Might it be efficient to construct a single paperclip factory? Then everything has a common structure and is easily maintained and repaired.

Contribute!

Comments URL: https://news.ycombinator.com/item?id=47066648

Points: 1

# Comments: 1

Categories: Hacker News

Show HN: UltraPlot 2.0 – semantic legends, better layouts, faster imports

Hacker News - Wed, 02/18/2026 - 3:41pm

UltraPlot v2.0.1 is out!

UltraPlot is a Matplotlib wrapper aimed at keeping Matplotlib’s flexibility while making common plotting workflows faster and more consistent.

v2.x focuses on semantic legends (categorical/numeric/size/geo), more reliable layout + axis-sharing in complex grids, guide architecture cleanup, CI hardening, and much faster import times via lazy loading.

We also launched a new docs site with a gallery: https://ultraplot.readthedocs.io/

Code: https://github.com/Ultraplot/UltraPlot

Feedback is very welcome, especially on legend API ergonomics and layout behavior in real figures.

Comments URL: https://news.ycombinator.com/item?id=47066090

Points: 2

# Comments: 0

Categories: Hacker News

Show HN: Sports-skills.sh – sports data connectors for AI agents

Hacker News - Wed, 02/18/2026 - 3:40pm

We built this because every sports AI demo uses fake data or locks you behind an enterprise API contract.

sports-skills gives your agent real sports data with one install command. No API keys. No accounts. For personal use.

Eight connectors out of the box: NFL, soccer across 13 leagues with xG, Formula 1 lap and pit data, NBA, WNBA, Polymarket, Kalshi, and a sports news aggregator pulling from BBC/ESPN/The Athletic.

npx skills add machina-sports/sports-skills

Open for contributions.

Comments URL: https://news.ycombinator.com/item?id=47066082

Points: 2

# Comments: 0

Categories: Hacker News

Pages