Software Development

3 of the 68 write-ups here are tagged Software Development. Every figure quoted below is in the figures table with the sentence it came from.

The figures

• 40% — 40% of review tasks needed manual backfill, costing about 15 extra minutes each time. →
• 1,000 tokens — Its original system prompt and tool descriptions came to under 1,000 tokens. →
• 1000-token — Still, I’d say the Pi base framework’s 1000-token limit seems overstated. →
• 1000 tokens — Its initial system prompt and tool description total under 1000 tokens. →
• $1.43 — The $1.43 and the $9.05 are both frontier models doing a job they were not specifically built for. →
• $1.43 — A single run costs about $1.43 on the first and $9.05 on the second. →

All figures, 644 rows

The write-ups

Your AI Didn’t Misread Your Code by Accident. You Handed It the Wrong Context.

“If your AI keeps writing bad code, you need a better model.” In each of the rebuilds below, a new model wouldn’t have fixed anything, because the model was reading the wrong thing.

reply box · telegra.ph

Stop Doing Manual DevOps: How I Use /loop and /hook to Automate My Daily Indie Hacker Tasks

As a solo developer shipping products alone, manual DevOps and repetitive data processing tasks are the silent killers of your side-project momentum, but configuring raw AI agents often creates mor…

reply box · telegra.ph

The Two Best AI Code Reviewers Score the Same. One Costs $1.43 a Run, the Other $9.05.

On ReactBench, GPT 5.6 Sol and Fable 5 posted Pass@1 scores of 43.1% and 41.2%.

reply box · telegra.ph

---

All 68 write-ups

---

Did this answer it? A star on the repository is the whole ask — it is what puts these in front of the next person looking; the data is CC BY and does not require starring.

Want a figure that is not here yet? Say which metric, which provider, which unit — one required field, and the page you came from is already filled in.

Got a better number? Open an issue — corrections and counter-data are the point.