Insights for running a business that doesn’t need you.
Practical operations and AI insights to help your business become easier to run, and less dependent on you.
Page 6 of 7
The End of the Tokeniser: Why Byte-Level Models Matter for Business Automation
Meta's Byte Latent Transformer removes the tokeniser and matched Llama 3 with up to 50% fewer inference FLOPs; a 2026 follow-up halved generation cost again.
5 min read
The Decision Log That Beats a Polished Data Room
A short written record made at the moment of each significant decision turns founder judgement into evidence a buyer can verify, while unrecorded reasoning stays unverifiable.
7 min read
OpenAI's Agents API Puts the Codex Harness in the Cloud, and the Data Rules Come With It
OpenAI's new Agents API runs the Codex harness as a managed cloud service: durable sessions, compaction and subagents with no platform fee. The catch is US-only data residency and no Zero Data Retention.
5 min read
Alibaba Open Code Review: When the Harness Beats the Model
Alibaba open-sourced its internal AI code reviewer. Its benchmark shows a structured harness more than doubling a general-purpose agent on the same model, at a ninth of the tokens.
5 min read
Anthropic's 2030 Economy Scenarios: Half the Economy Grows, Half the Payroll Is Exposed
Anthropic modelled AI's effect on the US economy to 2030: disruption concentrates in screen-based knowledge work, while physical-work wages rise in every scenario.
8 min read
The First Baseline Isn't Financial: Where Your New Team's Hours Actually Go
Before redesigning any workflow in a newly acquired business, run a two-week real-time time audit. It produces the measured baseline every first-100-days decision depends on.
10 min read
Google's Procedural Graphs Give AI Agents an Editable Answer to What To Do Next
Google's Procedural Graph paper gives AI agents an editable, self-evolving record of what to do next. Its agents ranked first in 21 of 24 benchmark settings, and the graph improved itself without retraining.
6 min read
What the Evidence Says AI Changes in the First 100 Days, and Who It Changes It For
The strongest workplace evidence finds AI assistance lifts output most for the least experienced staff and barely at all for experts. In the first 100 days that means measuring one candidate workflow before deploying anything, and writing the result into an automation candidate register.
9 min read
The Week-One Cash Reality Check: Can the Business You Just Bought Pay Its Own Way?
Three numbers in week one tell a new owner whether the business can pay its own way: the aged debtor report, actual days to payment against stated terms, and weeks of operating cover.
9 min read