This week: a new OpenAI model for professional work, a tool to automate Google Workspace with agents and data to choose cloud machines better.
GPT-5.4 brings reasoning, coding and computer use into one model
OpenAI released GPT-5.4 in ChatGPT, the API and Codex, along with GPT-5.4 Pro. According to the company, it is its first general-purpose model with native computer use, operating applications to carry out complete workflows, with up to 1 million tokens of context. On GDPval, which measures knowledge work across 44 occupations, it matched or beat industry professionals in 83% of comparisons, and it uses far fewer tokens to solve problems than GPT-5.2.
Tasks that involve spreadsheets, presentations and several systems at once become easier to automate with agents. Fewer tokens also lower costs for those already using the API in production.
How Dexpert helps: Artificial IntelligenceSources
- OpenAI: Introducing GPT-5.4
One command line for Drive, Gmail, Calendar, Sheets and all of Google Workspace
gws is a command-line tool for every Google Workspace API, such as Drive, Gmail, Calendar, Sheets, Docs and Admin, built for people and AI agents. It builds its commands by reading Google's own discovery service, so new API features show up automatically, and returns structured JSON. The project includes more than 40 ready-made agent skills, but warns that it is not an officially supported Google product and may still change a lot.
Companies that run on Google Workspace get a simple way to automate routines such as reports, calendars and file management with scripts or agents. For critical processes, wait for version 1.0 or use the official APIs directly.
How Dexpert helps: Digital TransformationSources
- GitHub: Google Workspace CLI
A comparison of 44 VM types across 7 clouds shows how much performance each dollar buys
Dimitrios Kechagias published his 2026 comparison of virtual machines across seven providers, including AWS, with 44 instance types tested in several regions. It includes new processors such as AMD EPYC Turin, Intel Granite Rapids and the ARM-based Google Axion, Azure Cobalt 100 and AmpereOne M. One of the main tips is to avoid older CPU generations, which tend to cost more and deliver less, and the author shows that performance varies even across regions of the same provider.
Switching instance family, without changing code, can lower the cloud bill or improve performance. Review the machines in use periodically and test the newer generations, including ARM ones.
How Dexpert helps: Cloud & ServerlessSources
- Dimitrios Kechagias: Cloud VM benchmarks 2026: performance / price
Want to put one of these to work in your company?
Dexpert builds custom systems, apps, integrations and AI solutions. Tell us what you need and get an estimate, or take the free digital maturity assessment.