5
IDEAS DAILY 1 SHIPPED BEFORE SUNDOWN
Project complete: 115 ideas sparked, 23 prototypes shipped.
2026-09-14
DAY 5 OF 23 RANK #8 WEBAPP
Theme: House

House Statistics: Property Intelligence & Home Search Tracker

UK property intelligence, live portal status verification (Active / Sold STC / Under Offer), admin criteria evaluation with custom important places, and an open heuristic valuation engine.

⚡ The Prototype // Idea #5

House Statistics: Property Intelligence & Home Search Tracker

Built in 4.5h with Antigravity, dlt, DuckDB, Python 3.13, Rich, httpx, curl_cffi, OSRM

tips_and_updates The 5 Morning Sparks

Someone's home, with roof and a few walls
IDEA #1 lightbulb

House Plan from Photos

Democratising house plan generation

I think this is a difficult project to pull off, but there must be a point at which adding a bit more easy to acquire data allows for creating a decent plan, and combining it with a refinement editor to Polish it off.

AI, Architecture
IDEA #2 lightbulb

AI Home redesigner

Edit your home with AI

Either photos or a 3d render of a home from Photos (I've seen someone do this in Astra), combined with AI which can change elements of the environment the user selects.

AI, Architecture
IDEA #3 lightbulb

Smart home inventory

Control your smart home from a smart phone

There is an opportunity for AI in creating skills for the smart home, but the first step is registering the devices and creating a database for them. This idea exists already, I am sure, and this would be an opinionated clone that makes it AI first.

Tool, Hardware, Data
IDEA #4 lightbulb

Campter Van fitouts

Inspiration database for Vanlifers

Create a scraper for Camper Van photos to inspire and allow exploring. It could be linked to instructions, descriptions, etc and have a vector database to make it a really powerful search engine. Maximum amount of meta tagging and good taxonomies are welcome

Data, Travel
IDEA #5 lightbulb

House statistics

Real estate scraper to feed a home search

A lot of websites repurpose available statistics to sell a product for home buyers. But they don't do it the way I would prefer, even if I pay. So if I am interested in a specific home, can I check it out on property websites, council records, criminal records, school, environmental, etc. to produce a report for me?

Data, Finance SHIPPED
construction

How It Was Built With AI in 4.5 Hours

1

Step 1

Chose Idea 5 and fleshed out the spec around scraping and free APIs

2

Step 2

Decided on a standalone repo, dlt, duckdb, and ad hoc portal checks instead of mass scraping

3

Step 3

Spun up repo, fixed 3.14 segfaults, and added OSRM walking distance and crime per capita

4

Step 4

Fixed Cloudflare blocks with browser TLS impersonation and extracted full listing metadata

5

Step 5

Added home search portfolio tracker, market status verification, criteria scoring, and heuristic valuation

thumb_up

What Rocked

  • ✓It span out a new codebase almost instantly, with a decent first pass
  • ✓Open heuristic valuation and live URL metadata auto-population worked smoothly without heavy dependencies
bug_report

What Broke & Fixed

  • ✗Multiple turns over dlt, not actually using a good structure for dlt
  • ✗Zoopla Cloudflare anti-bot checks blocked standard requests until switching to TLS impersonation
smart_toy AI Interaction Summary
Genesis Prompt: "/spec-implementation"
Interaction Turns: 16 human turns across 1521 execution steps

Key Guidance Turns:
  • Turn 2: python 3.14 please, check the pages on zoopla, rightmove and onthemarket as well (ad hoc, not scarping) Create a new repo for t...
  • Turn 3: yes
  • Turn 4: I would like to be able to generate a report from a zoopla/onthemarket/rightmove link, so it would pull the data and postcode f...
  • Turn 5: 1. [instruction for AI: Record the steps from AI in short, casual tone] 2. It span out a new codebase almost instantly, with a ...
  • Turn 6: decrese steps to 4 and make it based on conversational turns, not on agent implementation steps What broke: Multiple turns over...
  • Turn 7: ok remove the fluff from steps. Just 4 untitled items, it sounds 7/10 like AI, make it sound 0/10 like AI and more like how oth...
  • Turn 8: The tool doesn't work with zoopla, because of Cloudflare https://www.zoopla.co.uk/for-sale/details/74208604/?search_identifier=...
  • Turn 9: I don't think it works in generating real results for new queries. I queried the above and it comes back with Wesminster City S...
  • Turn 10: I can see that the results are not quite right. It misses at least one school - St Cathrine's. Are we getting the data from the...
  • Turn 11: extract more infor from the listing. Info like the size in sq feet or sqm, main image link, leasehold information, any other me...
  • Turn 12: I don't think the shops are right either. They should have more shops, including big ones that don't appear before ones far away
  • Turn 13: I don't think the store search is working as intended. For example this property https://www.zoopla.co.uk/for-sale/details/7263...
  • Turn 14: all right. I want to use this for a home search in the future. I want to be able to manage which properties I am tracking, and ...
  • Turn 15: can I add things for tracking by URL, which then populates the metadata for the tracked property?
  • Turn 16: update today's entry

Tools Executed: run_command (349), view_file (199), replace_file_content (58), write_to_file (50), manage_task (30), grep_search (16)
Files Modified: .gitignore, EN2_7BT.md, EN5_2EX.md, README.md, SKILL.md, SW1A_1AA.md, SW1E_6AL.md, __init__.py