Monday, October 19, 2026Radar run of Oct 19: 25 new entries, 5 read in full, 20 quotations checkedabout 5 minutes

Anthropic reports from its own records that junior technical hiring fell from 27% to 9% as agent hours reached six times employees' hours.

How to read this pageSerif their words, as the source published themMarked found word for word in the sourceSans our words

Oct 18
Read asSame facts, in the order each reader needs them.

What changed today

Decisions, new evidence, corrections and commitments we read in full, one per story. Arguments and recaps don’t count.

New evidenceAnthropic

Anthropic reports from its own records that junior technical hiring fell from 27% to 9% as agent hours reached six times employees' hours.

Research: How employment is changing at Anthropic

What it means for youAnthropic reports that it now hires far fewer junior technical staff than a year ago, while its AI agents work six times as many hours as its employees.

For lawmakersAnthropic's authors say company measures cannot offset AI job disruption and call for government retraining and income support, guided by a clear picture of the scale of change.

For researchersA within-person comparison of 192 adopters with 74 non-adopters finds output rising after adoption, with the authors noting that adoption is not random and may bias estimates.

“Figure 1 shows that the share of junior-level employees in technical departments at Anthropic slipped from 21% to 14% since early 2025. The proportion of junior technical hires fell even faster, from 27% to 9%.”

The report's central hiring finding, based on Anthropic's internal employment data.

AnthropicAnthropic, Oct 16 · found word for word

Said today

One checked quotation per person, people with the power to act or first-hand knowledge first, at most two from any one group.

Who is quoted here: 2 with no group recorded yet.

Group not yet recorded

“We argue that economic development in the countries that we serve and partner with that could have taken centuries to make happen, the potential is there for that economic development now to happen within decades”

Donohoe, speaking to The Guardian at the World Bank and IMF meetings in Bangkok, on what small AI could do for developing economies.

Paschal Donohoechief knowledge officer, World Bankquoted second-hand · The Guardian, Oct 18 · found word for word

Group not yet recorded

“You need access to electricity. You need access to digital infrastructure. People have to be geared to adoption. And the government should be able to give some safety, protect against bad actors so that adoption can work.”

Georgieva on a Monday panel at the Bangkok annual meetings, on what governments must provide for AI adoption.

Kristalina GeorgievaManaging Director, IMFquoted second-hand · The Guardian, Oct 18 · found word for word

Watch and listen

Interviews and talks we read in full, by what they add, then by reach. ▶ opens the video at the moment quoted.

1 of 7 videos and podcasts in this run were read in full, from captions or a transcript. Only those are ranked here; the rest are under “Everything else”, from their descriptions.

New research

Papers and reports, the ones we read in full first. The finding in our words, how it was reached, and what it cannot show.

5 of 5 research items.

Company reportAnthropicread in full

How employment is changing at Anthropic

Found
Junior-level share of technical staff slipped from 21% to 14% since early 2025 and junior technical hires fell from 27% to 9%. By August interactive agent hours were six times employee hours; code commits per engineer rose to 5x six months earlier.
How
Analysis of Anthropic's internal data on employment, AI use, output, surveys, communication, calendars and hiring records, plus 17 employee interviews; within-person comparison of 192 engineers adopting coding agents with 74 who did not.
Limits
Single company, with revenue growth and stock ownership that the authors say cannot be replicated across the economy. Tool adoption is not random and may bias estimates; more code is not necessarily better; agent-adopter results partly reflect prior upward trends.

ReportLessWrongfrom the description only

More Capable Models Write Better Probe Evasion Prompts

Found
Per the post's title, more capable models write better probe-evasion prompts; the excerpt gives no numbers.
How
Probes were trained on Qwen3.5-27B (no reasoning) activations on answer tokens for arithmetic, ordering, belief tracking and contradiction tasks; 14 models then each wrote evasion prompts over 16 iterations of probe-score feedback.
Limits
The authors describe it as a two-week sprint project; further limits are not in the excerpt.

ReportLessWrongfrom the description only

LLMs Learn to Evade Latent Monitors from Prior Feedback Alone

Found
The authors report the models' own edits are too small to evade the monitor, but scaling them up linearly induces evasion, and the trained LoRA evades monitors for concepts it never saw during training.
How
Off-the-shelf language models were given a latent monitor's verdicts on their earlier responses and asked to edit their activations in context; a rank-1 LoRA was then trained to apply a scaled-up edit within the forward pass.
Limits
Not stated in the excerpt; the full post and paper were not read.

ReportLessWrongfrom the description only

On Qwen, the J-lens readouts at the newline token reveal lies

Found
The author reports a lie-related word family at the newline token after 84% of Qwen's lies and 9% of truths (78% and 7% held out); eight families as a detector caught 86% of held-out lies and wrongly flagged 13% of truths (AUROC 0.87); probes there reached held-out AUROC 0.96 against 0.69 to 0.81 elsewhere.
How
Four models (Qwen3.6-27B, two Gemma-3 models, gpt-oss-20b) were placed in MASK pressure scenarios that give a role and a reason to lie; readouts at chat-template tokens were grouped into word families and counted after lies and after truths.
Limits
Not stated in the excerpt; the full post was not read.

ReportLessWrongfrom the description only

How Researchers Lie to AIs

Found
The excerpt shows only that a figure gives the percentage of evaluations containing each lie type; the percentages are not in the excerpt.
How
A survey of 292 public alignment evaluations published between December 2022 and October 2026, classifying the type of lie each told the model.
Limits
Described as a two-week sprint project; the harms of lying are discussed as plausible, not measured, in the excerpt.

Stories that continued

Events several sources covered, read as one thread across days.

Coming up

Dates when a decision can still change, from our decision briefs, and how to take part.

  1. 26Oct

    United Kingdom

    Lords Report stage of the Cyber Security and Resilience Bill, in the Lords Chamber: the next chance to amend it. The list of amendments is still open until the day.

    Should the UK make reporting serious AI incidents a legal duty?

Questions you could put to a minister

  1. Will the Government table an amendment on reporting serious AI incidents before the Lords consider the Bill on October 26? If not, why not?
  2. If AI incident reporting is to be required, who should have to report: the developer of the AI system, the organization using it, or both?
  3. What should count as a reportable incident or near miss, and who should receive the reports and act on them?

From the decision brief, in neutral words. We recommend no answer.

Where the answers stand

2 of the nine questions have an answer we keep up to date, with the reason for every change. All nine questions

Everything else today: 19 more entries, filed under the nine questions

Can AI systems act beyond what people intended or allowed?

Are AI companies putting stronger safeguards into practice?

Who can demand evidence, delay a release or stop an unsafe deployment?

How is AI used in attacks, defence and war?

What happens to work, income and the economy?

How is AI changing childhood, health and daily life?

How fast is AI advancing, and where is it heading?

Search and filter the week’s record

How this edition was made

So you can judge how far to trust it.

  1. 113 of 114sources answered
  2. 288items they published, October 18–19, 2026
  3. 32kept by a wide keyword filter
  4. 25became entries, 5 read in full
  5. 20quotations found word for word in the saved text
How this edition chooses
  • What changed: entries whose change is a decision, new evidence, a correction or a commitment, that we read in full; most important first; one per story.
  • Said: one checked quotation per person from the entries that matter, people with the power to act or first-hand knowledge first, at most two from any one group.
  • Watch and listen: interviews and talks we read in full, by importance, then by views.
  • Decided and promised: decisions, corrections and commitments, decisions first.
  • New research: papers and reports, the ones we read in full first.
  • The headline: written for the edition when we can; otherwise the top change.

Agents pick the items, write the summaries and the “why it matters” lines, and label what kind of change each one is. Code checks every quotation against the text we saved: the words must be there, in order, without a dropped “not”. That proves the words are in the source; it does not prove the person meant them as quoted. English sources only; X is not read yet.