AI promised less work. Oioannou has never worked more – Sulat

A report from Oioannou puts AI promised less work. Oioannou has never worked more on the technical radar. This concise read preserves the source’s central claim, its context and the questions worth watching next.

What the source reports

Agents made implementation cheap, so Oioannou started doing far more of it in parallel. Oioannou’s GitHub history since 2019 shows what that did: as VP of Technology Oioannou is back at Oioannou’s full-stack engineer pace, with bursts, burnout and a low season that is no longer low.

Original article image for AI promised less work. Oioannou has never worked more

The broader context

Agent systems are defined by more than the underlying model. Tool access, retries, memory, handoffs and approval boundaries determine whether a workflow remains useful when conditions change. A polished demonstration can hide the operational work required to supervise long-running tasks and recover from partial failures.

A practical agent review follows a task from assignment through recovery. Record where the agent found context, which tools it called, how often it asked for help and what happened when a tool failed or returned stale information. Test an ordinary task and one with incomplete or contradictory inputs. Count silent errors as well as hard failures. The useful metric is not how long an agent can run, but how reliably it finishes with an auditable result and an appropriate amount of supervision.

A compressed source summary cannot carry every operational detail. It may omit rollout timing, benchmark construction, eligibility, dependencies or conflicting results. That does not make the source wrong; it defines the limit of what the summary alone proves. Treat the initial report as a claim to be tested rather than a universal outcome. Check availability and version details for releases, reproducibility for research, and delivered usage for business claims. Where possible, run a small reversible evaluation and keep its measurements. Corrections and follow-up reporting can materially change the interpretation, especially when the underlying technology, pricing or policy is moving quickly. This distinction between a publisher’s evidence and its broader applicability is essential to responsible technical coverage.

For readers evaluating AI promised less work. Oioannou has never worked more, the publisher’s account is the starting evidence. Independent implementation details, subsequent corrections and real-world outcomes provide the test of how broadly the claim applies.

Why it matters

The important question is how the system behaves under supervision: what tools it can use, what context it retains and how people detect and recover from mistakes.

How to assess the claim

Test permissions and failure recovery before judging autonomy.
Measure human intervention across long tasks, not only successful demonstrations.
Review logs, tool boundaries and the agent’s access to sensitive context.
Check whether evaluations include changing conditions and incomplete information.