“I’m tired of reading people talk about how we’re “in the era of agents” that don’t fucking work and will never fucking work.”
Our reading, the sentence that was checked:
AI agents do not work and will never work.
Assessment: Contradicted Settled before its deadline on 2026-10-29 Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, The Hater's Guide To The AI Bubble, paragraph 16.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-the-haters-gui-8c577b-a2
We recorded 49 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-07-21. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Forecast
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- Settled before its deadline
- Deadline state on 2026-10-29
- Settled before its deadline (the release’s reference date)
From the speaker
- agents
- don’t fucking work
- will never fucking work
Our interpretation
- The ordinary reading is that AI agents cannot become practically useful at carrying out their intended multistep tasks.
- A bounded agent can work without being a universal worker. Human setup and review are compatible with useful autonomous execution of the intervening task.
- A single selected demonstration is insufficient. Repeated useful results, verified outside the model developer, can contradict the categorical future claim.
- The present-tense component concerns publication, but the supplied forecast classification makes the categorical never component the main adjudication target.
Counts as supported if: No finite period of failures establishes that agents will never work. Supporting that categorical prediction would require a demonstrated general barrier that rules out useful future agent execution.
Counts as contradicted if: A deployed agent repeatedly performs useful multistep work, with its results independently verified by the organization using them.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
Outcome: Contradicted
The dated evidence goes against what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- Settled before its deadline
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
Anthropic's March 6, 2026 account describes February Firefox vulnerability discovery using iterative agent execution and tool feedback, with human validation [e1]. On April 21, 2026, Mozilla confirmed that the Opus 4.6 collaboration led to fixes for 22 security-sensitive bugs in Firefox 148 [e2]. The documented multistep discovery workflow and repeated findings confirmed and fixed by the software maintainer establish practical usefulness in a real security workflow [e1, e2]. This meets the early-resolution rule against the categorical prediction that agents will never work [e1, e2].
Limitations
- Mozilla verified the results as the software maintainer but was also Anthropic's collaboration partner [e2].
- The disclosures do not provide a complete per-attempt success rate or establish reliability across unrelated tasks [e1, e2].
- The workflow included human validation and maintainer involvement; the evidence does not establish fully autonomous discovery through remediation [e1, e2].
- Production-agent research identifies reliability and substantial human oversight as continuing constraints [e3].
- These later results do not settle the present-tense assertion about agents on July 21, 2025.
Evidence (3 items)
Goes against the claim: Anthropic,
real-time feedback
Why it matters here: Anthropic describes iterative tool use in discovering Firefox vulnerabilities, with human validation and Mozilla's involvement. This establishes an agent workflow rather than a one-response chatbot demonstration, although it is developer-reported evidence.
- Publisher
- Anthropic
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Goes against the claim: Mozilla,
fixes for 271 vulnerabilities
Why it matters here: Mozilla confirms that its evaluation of Claude Mythos Preview produced many findings that were verified and fixed in a released browser. Together with the documented discovery workflow, this establishes repeated practical usefulness in a bounded task.
- Publisher
- Mozilla
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: arXiv,
the top development challenge
Why it matters here: This study of production deployments identifies reliability and substantial human involvement as continuing constraints. It limits any broader conclusion about autonomous agents, but does not negate the verified security use case or establish permanent impossibility.
- Where
- Abstract
- Publisher
- arXiv
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, The Hater's Guide To The AI Bubble
I’m tired of reading people talk about how we’re “in the era of agents” that don’t fucking work and will never fucking work.
- Where
- paragraph 16
- Published
- 2025-07-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-the-haters-gui-8c577b-a2: Current. Outcome: Contradicted. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
HeldBy this site
The final-text check failed: conditions_shown.
About assessment wyea-the-haters-gui-8c577b-a1, which was not published.