Opus 5.5, Sonnet 5.5, and GPT-6 make intelligence cheaper to use
Anthropic and OpenAI are lowering the cost of useful work. Token prices explain part of it; efficiency, model routing, and verification explain the rest.
Lisbon · online
Mário Salgado
I lead teams building AI systems that survive contact with production. As Head of Data & AI at QuintoAndar, I work on the conversational platform, where my days revolve around LLMs, agentic systems, and the engineering required to make them reliable at scale.
$ whoami
mário — engineering leader · systems builder
$ cat ./focus.txt
data platforms · agentic AI · production engineering
$ ls ./interests
agents/ architecture/ data-platforms/ open-systems/
$
Tech Stack
- LanguagesPython · TypeScript
- AI & agentsGoogle ADK · MCP
- Data & cloudData platforms · AWS · GCP
- ProductionEvals · Observability
Latest from the blog
all posts →Ultrafast makes GPT-5.6 Sol feel like a different product
OpenAI's Ultrafast mode does more than shorten response times — it changes how often I iterate, how early I correct direction, and how many alternatives I explore.
Trying Omarchy 4.0: why Mac users are suddenly Linux-curious
Omarchy Quattro packages Arch, Hyprland and an agent-friendly desktop into something Mac users can understand — and DHH is turning that curiosity into a movement.
Cheaper tokens will not mean smaller AI budgets
OpenAI cut GPT-5.6 Luna prices by 80% and Terra by 20%. For big AI customers, lower unit costs are more likely to expand consumption than shrink the forecast.