Phroneses.com

The recent reporting on two unauthorised computer breaches, controlled by AI agents, highlights the challenges posed when using LLMs to specify the goal for an agent.

Humans are bad at estimating.

The approach to AI in the US and China is in stark contrast.

Generative large language models excel at creating new items based on statistical patterns.

This article quantifies the difference between human-authored and agentic-produced software.

Are you a problem solver or a problem explainer?

When using an LLM to modify pre-existing code, it is not possible to get the error rate to zero.

State-driven and history-interpreting distributed system algorithms accumulate state into a local data structure and make decisions by interpreting that structure.

In a traditional factory, machines are used to build products.

A challenge when building RESTful services is trusting the data sent from the client, especially if the service uses that data to make processing decisions.

In a discussion on technical debt, the point often made by engineering is "if we do not pay attention to the debt right now, in the future, we will not be able to effectively engineer as the debt will be too great".

When calling a remote service in a distributed system, it is never possible for the client to know if the operation was successfully applied.

A residual design focuses on the present in a way that does not curtail the future.

Many engineering leaders say "my engineers all use AI but we are not shipping any faster."

Your CEO and board have asked for a strategy and you have three weeks to deliver it.

AI has increased delivery speed, but reliability becomes opaque without measurement.

Specification driven development (SDD) encourages a "complete" spec written up-front in the hope that the detail can be used to more accurately drive AI agentic code generation.

Artificial intelligence now appears in many areas of daily life.

LLM use can get expensive if the number of per‑call tokens is unchecked.

Asynchronous IO does more than overlap work.

Hiring must shift from evaluating code production to evaluating engineering judgment because AI has collapsed the cost of typing but raised the cost of misunderstanding.

We have reached the end of major gains from large language models (LLMs).

Vibe coding produces code.

The engineering landscape has shifted.

Leaders often find themselves in a situation where urgency keeps increasing but progress does not follow.

AI is already reshaping your delivery workflows, whether you see it or not.

LLMs can generate correct code, but they cannot safely change real software systems.

Getting software delivered has always required a specification.

Modern software delivery behaves less like a pipeline and more like a network: intent, decisions, and constraints move between team members.

AI in 2026 demands more from you than simple instructions.

This is a shorter, more general version of the original article which focuses on how software delivery occurs and how Team AI can unleash more benefits.

Large language models such as GPT‑5.4, Claude Opus 4.6, and DeepSeek R1 are now everyday tools.

Individual AI has already made engineers faster.

Individual AI has already made engineers faster.

Artificial intelligence has entered a new phase.

AI is now infrastructure.

Luxury watchmaking faces pressure to adopt AI at the pace of mass‑market retail, yet most AI trends undermine the very qualities that define a maison: scarcity, discretion, and narrative integrity.

Artificial intelligence is reshaping how organisations operate, communicate, and compete.

define its real capabilities and limits.

AI adoption is no longer a technical experiment.

AI does not fit into business‑as‑usual software delivery.

We interact with AI systems through natural language.

Most latency comes from retrieval hops, long prompts, and serial tool calls.

Recent headlines have shown the same unsettling pattern.

Many people now use chat systems powered by artificial intelligence for writing, research, planning, or quick explanations.

AI systems behave like probabilistic components, so evaluation must focus on observable behaviour rather than idealised test cases.

Large language models have introduced a profound shift in how software systems are conceived, built, and governed.

Artificial intelligence is a practical tool that speeds up routine thinking tasks.

Large language models behave like probabilistic components, not deterministic functions.

This guide gives beginners a clear, practical foundation for working with AI chatbots.