PUBLIC SIGNAL OBSERVATORY

AGENT SMITH INDEX / 01

Capabilities / infrastructure / markets / public interest

Connecting …

               .:iiiiiiiii:.
            .:iillllllllllli:.
          .:illlllllllllllllli:.
         :illlllllllllllllllllli:
        llllll:'           ':lllll
        lllll:               :llll
        llll:                 :lll
        :lll  ...         ...  lll:
        .lll .::::.     .::::. lll.
        :ll=====================ll:
        :l|@@@@@@@@|=|@@@@@@@@|l:
         :|@@@@@@@@| |@@@@@@@@|:
          ':@@@@@@:' | ':@@@@@:'
          |         .|.       |
          |        :_|_:      |
          :       .     .     :
           |    .:-------:.  |
           :      '-----'    :
            \      ...     /
             '.          .'
             /:'-.____.-':\
         _.-/ |          | \-._
     _.-'  /  |\        /|  \  '-._
 _.-'     /   | '.    .' |   \     '-._
/        /    |   >--<   |    \        \
        /     |  /####\  |     \
       /___   | /######\ |   ___\
           /  |/  ####  \|  \
          /   |   ####   |   \
         /    |  ######  |    \
        /     |  ######  |     \

Sources online–Status 15 min · news 1 h · websites 6 h

Open operational incidents–Not evidence of AGI or an attack

News topic matches · 48 h–Keyword matches, not yet verified

Stored observations–Building a baseline

01 / NEWS PULSE

48 hours · 2-hour bins

All stored publications in the last 48 hours, independent of the report list limit. Duplicate coverage may remain.

HOW TO READ THIS

Evidence behind the clock

Match: relevant text found.
Measurements: data available, threshold assessment pending.
No data: not currently measurable.

A supported finding requires suitable measurements, independent corroboration and assessment of alternative explanations. This verification is not yet automated.

These working criteria are not a scientifically validated AGI scale. Policy reactions are not a prerequisite.

02 / LAB WATCH

Source status · METR coverage

03 / SIGNAL FEED

View all ↗

Sensor network

Coverage details ↗

04 / AI 2027 REALITY CHECK

Read AI 2027 ↗

1 observed · 1 partial · 3 not assessed 5 selected milestones · reviewed 15 Sep 2026

2025Coding agents in use● Observed

Early 2026AI accelerates AI R&D◐ Partial evidence

2027 scenarioSuperhuman coder○ Not assessed

2027 scenarioSuperhuman AI researcher○ Not assessed

2027 race endingGeneral superintelligence○ Not assessed

Evidence & assessment limits

Coding agents: Anthropic documents real use of Claude Code for delegated software tasks. This supports adoption of coding agents, not every detail of the scenario's 2025 predictions. Provider study ↗

AI-assisted R&D: Google reports AlphaEvolve improvements to AI training and algorithms. These are specific provider-reported gains, not independent verification of the scenario's company-wide 50% research acceleration. AlphaEvolve evidence ↗

Later milestones: Not assessed in this initial review. This does not mean they are disproven or known to be absent. Matching the scenario's definitions requires a separate evidence review. Milestone definitions ↗

Editorial snapshot of five selected milestones, not a completion percentage for all AI 2027 predictions. Statuses are not updated automatically from news keywords. Dates refer to the original April 2025 scenario; the authors have since linked updated forecasts. The community clock and automatic alerts are separate.

05 / JOBS & AI

US labour releases ↗

Employment releases and AI-related reports. Job exposure, job losses and unemployment are different measures; AI causation is not established by keyword matches.

Loading publications…

Observations

Fetching initial sources …

Benchmarks → companies

METR model names are matched to companies using an explicit alias list. Unknown models remain unmatched. Other providers are tracked as an additional watchlist.

View METR measurements

DATA NETWORK / COVERAGE

Collection health at a glance. Successful fetches do not imply complete coverage of a topic.

Sources & coverage

Not yet connected

X counts, sampled posts and trends, TomTom road flow and daily market data are connected. Reddit, company website traffic and direct office activity remain pending. ARC/SWE page text is not a structured leaderboard. Collectors use public sources and configured API access.

Learning history

Reports are prioritized using transparent text rules. You may optionally correct an assessment. Reviews and observed changes are retained with timestamps.

Rule version: . Topic weights are not yet adjusted automatically. Ongoing collection builds the baseline; historical imports are not retrospective predictions.