Practical notes from the Conceptdigital team on building software: infrastructure, product decisions and delivery.
25 October 20265 min read
In 2008 the Obama campaign tested 24 versions of its sign-up page. A "Learn More" button and a family photo lifted sign-ups by 40.6%, worth an estimated $60 million. Here is why a smaller ask wins, and how to test it on your own site.
17 October 20266 min read
Blocking IPs and locking accounts both fail against modern login attacks. The fix is to score each request cheaply and only spend the expensive password check on traffic that looks real.
15 October 20266 min read
A developer cut their app's start-up on a bad connection from 35 seconds to 3 by trusting the saved login, caching the home screen and running requests in parallel. Here is why it works, what it costs, and how to test it.
15 October 20268 min read
At a McKinsey alumni panel, two chief AI officers explained why most companies use AI without transforming the business: three gaps between pilot and P&L, a portfolio of anchor use cases, and ROI tracked inside the tool.
30 September 202613 min read
A viral reel says six gaps in AI-built apps add up to $100K of liability before the first sale. We checked each against the law. Most are overstated; all are cheap to fix.
29 September 202612 min read
Daniel Walsh ran 13 AI decision model set-ups through 7,671 questions. Cloud-hosted Jev led overall at 95.23%, but local models came within a point and won some tasks.
28 September 20269 min read
The same API on the same $12 server: Python held 2,150 concurrent users, Node about 3,500. What Arjay McCandless's test means for choosing a stack.
26 September 20269 min read
A $12 a month server held 5,000 concurrent users in Arjay McCandless's load test after two caching changes. What that means for your hosting budget.
