IndieFounder publication
Practical stories, product lessons, teardowns, and field notes for independent founders.
Startups
Exact caching is useful for deterministic or repeated AI work, while semantic caching needs stricter freshness rules. Version caches by prompt, model, a…
Archive
Startups
Exact caching is useful for deterministic or repeated AI work, while semantic caching needs stricter freshness rules. Version caches by prompt, model, a…
Startups
Route work by requirements instead of using one model everywhere. Classify tasks into capability tiers, validate cheap-model output, escalate on failure…
Startups
More context can increase cost and reduce signal. Separate current task state, durable memory, retrieved knowledge, tool results, and raw history; summa…
Startups
Syntactically valid JSON can still violate business rules. Treat model output as untrusted input: validate a schema, apply application-level policy, res…
Startups
Production reliability requires recovery, not just better prompts. Separate reversible and irreversible actions, persist checkpoints before side effects…
AI
Anthropic shipped Haiku 5.5 on 7 October at $0.10 / $0.50 per million tokens under 100k. Over that line the rate is 5x. The new tokenizer spends about 30% more tokens.
Startups
Use an agent only when uncertainty, tool choice, or adaptation creates real value. Fixed jobs such as scheduled exports, deterministic calculations, and…
Product & SaaS
Anthropic priced Claude Haiku 5.5 at $0.10 / $0.50 on 7 October 2026, and OpenAI's GPT-6 Luna sits on the same list-price band. Indie SaaS should shadow the high-volume routes, not flip the product.
Product & SaaS
Rewrite decisions should follow measurable constraints: change cost, failure frequency, architecture limits, and product opportunity cost.
Product & SaaS
A small SaaS needs a handful of useful signals: uptime, errors, latency, database health, queues, billing events, and customer-visible failures.