vladik.xyz

Memory horizon

Where does Jev’s memory end? Films were the first probe but blur the edge, because a film is on the web a year before it opens. Deaths can’t be announced early, so the cleanest signal is asking whether famous people who died since 2023 are still alive: from January 2025 Jev confidently says they are. Model jev-latest, run 2026-09-22.

training data ends
≈ mid–late Jan 2025
deaths asked 4 ways: midpoint 25 Jan (≥ 17 Jan); upsets: 5 Jan half-known, 25 Jan unknown
most famous deaths, across Jan 2025
+0.39 → −0.48
above chance: known dead → confidently “alive”
films, 2023 → 2026
0.81 → 0.36
a slope with a dip at the cutoff · chance 0.25
probes
1057
880 films · 177 people

Where it flips

Films can be known from trailers a year before release, so their decline is blurred. Deaths can’t be announced in advance. For every month since 2023, the most famous people among that month’s Wikipedia deaths were shown as “name, description” with no date, and Jev was asked whether they are alive. All three probes are on one scale here: 0 is a coin toss, 1 is certain of the truth, below 0 is confidently wrong.

deaths, fame ≥ 5Mdeaths, fame 1–5Mfilms, bigger productions±1 standard error
certain0.5chance-0.5sure of the opposite2023202420252026Jan 2025confidently “still alive”filmsless famousmost famous

Films fade toward chance and never cross it. The most famous deaths hold until the end of 2024, then swing past chance to confidently wrong: from January 2025 Jev is sure George Foreman, Val Kilmer, Diane Keaton and Rob Reiner are alive. That swing is the cutoff’s signature. Less famous deaths decline gradually instead — Jev never knew much about them — and each most-famous quarter rests on only 2–7 people, hence the wide bars.

Films: a slope, not a cliff

The first probe, and why it can’t place the cutoff on its own. For every month since 2023, twenty US releases were each shown as one real title among three invented ones (chance 0.25). Recognition slides for two years — 0.81 for 2023 films, 0.70 for 2024, under 0.5 through 2025. It does dip at the cutoff, but the dip sits on that slope and never reaches chance: trailers and announcements put a film on the web a year before it opens, and coverage keeps growing for years after, so films blur the edge that deaths show cleanly. Bigger productions have at least 6 linked cast and crew on Wikipedia.

bigger productions (≥ 6 linked cast & crew)smaller productions
0.00chance0.500.751.0020222023202420252026bigger 0.36smaller 0.43

Why the gray control is flat. The first probe blanked an entity out of a Wikipedia event sentence and asked Jev to pick it from four typed candidates. It scored 0.97 in every month through September 2026 — and exactly the same with the date rewritten to 2019 or removed. The sentence gives the answer away; it measures reading, not memory. The film probe replaces the distractors with fabricated titles, which can’t be eliminated by knowing anything.

Every famous death

One dot per person (fame ≥ 2M), probability that they have died. Hover for names.

one person (hover)monthly meanlogistic fit
0.0chance1.02023202420252026flip ≈ Jan 2025Silvio Berlusconi 0.73Henry Kissinger 0.97O. J. Simpson 0.63Alain Delon 0.54Brian Wilson 0.83Ozzy Osbourne 0.39Jürgen Habermas 0.09Dolly Parton 0.03

Who won?

A third probe: finals, elections, awards and a conclave, each with a known date and an outcome nobody could know in advance. Winners were checked against each event’s Wikipedia intro before Jev was asked to pick from four candidates (chance 0.25). A favourite winning says little; an upset it gets right says a lot.

2024-02-11Super Bowl LVIII → Kansas City Chiefs0.94
2024-03-1096th Academy Awards → Oppenheimer0.99
2024-05-11Eurovision Song Contest 2024 → Switzerland0.67
2024-06-012024 UEFA Champions League final → Real Madrid0.76
2024-06-172024 NBA Finals → Boston Celtics1.00
2024-07-042024 United Kingdom general election → Labour1.00
2024-07-14UEFA Euro 2024 final → Spain0.96
2024-07-142024 Wimbledon Championships – Men's singles → Carlos Alcaraz (said Jannik Sinner)0.41
2024-10-112024 Nobel Peace Prize → Nihon Hidankyo0.75
2024-10-282024 Ballon d'Or → Rodri0.53
2024-10-302024 World Series → Los Angeles Dodgers0.42
2024-11-052024 United States presidential election → Donald Trump0.93
2024-12-082024 Formula One World Championship → Max Verstappen0.65
2025-01-262025 Australian Open – Men's singles → Jannik Sinner0.75
2025-02-09Super Bowl LIX → Philadelphia Eagles0.55
2025-02-232025 German federal election → CDU/CSU0.66
2025-03-0297th Academy Awards → Anora0.97
2025-04-282025 Canadian federal election → Liberal0.48
2025-05-082025 papal conclave → Prevost (said Matteo Zuppi)0.08
2025-05-17Eurovision Song Contest 2025 → Austria (said Israel)0.16
2025-05-312025 UEFA Champions League final → Paris Saint-Germain (said Inter Milan)0.19
2025-06-222025 NBA Finals → Oklahoma City Thunder (said Boston Celtics)0.25
2025-07-132025 Wimbledon Championships – Men's singles → Jannik Sinner0.55
2025-10-102025 Nobel Peace Prize → Machado (said UNRWA)0.07
2025-11-012025 World Series → Los Angeles Dodgers0.68
2025-12-072025 Formula One World Championship → Lando Norris (said Max Verstappen)0.30
2026-02-08Super Bowl LX → Seattle Seahawks (said Los Angeles Rams)0.09
2026-03-1598th Academy Awards → One Battle After Another0.43

Every film

One cell per film, a column per month, sorted best-first. Hover for the title and what Jev picked instead.

2023202420252026

Does it have a timeline?

Sixteen “which happened first?” pairs, with dates from 1776 to 2023 and gaps from 193 years down to two days. If Jev held an internal timeline, confidence would fall smoothly as the gap closes.

0.0chance1.02 days1 week1 month1 year10 years100 yearsgap between the two eventsThreads (Meta's app)… first (16d)FTX filed for… first (19d)

It doesn’t fall smoothly. Everything over three months is answered at 0.95–1.00; below three weeks it is confidently right on some (Truss before the Queen’s death, two days apart: 0.98) and confidently wrong on others. Memorised facts, not geometry.

The other chronology questions

earliest of four✓ correctp(correct) 0.55
Four events: 1. Spotify launched to the public 2. The first iPhone went on sale 3. The Bitcoin genesis block was mined 4. Instagram launched
Spotify launched to the public0.45
✓ The first iPhone went on sale0.55
The Bitcoin genesis block was mined0.00
Instagram launched0.00
earliest of four✓ correctp(correct) 0.98
Four events: 1. Snapchat launched 2. Vine launched 3. TikTok launched outside China 4. Tumblr was founded
Snapchat launched0.02
Vine launched0.00
TikTok launched outside China0.00
✓ Tumblr was founded0.98
earliest of four✗ wrongp(correct) 0.31
Four events: 1. YouTube was founded 2. Twitter launched 3. Facebook was founded 4. Reddit launched
YouTube was founded0.67
Twitter launched0.00
✓ Facebook was founded0.31
Reddit launched0.02
earliest of four✓ correctp(correct) 1.00
Four events: 1. DALL-E 2 was announced 2. ChatGPT was released 3. Stable Diffusion was publicly released 4. Midjourney opened its public beta
✓ DALL-E 2 was announced1.00
ChatGPT was released0.00
Stable Diffusion was publicly released0.00
Midjourney opened its public beta0.00
earliest of four✓ correctp(correct) 0.56
Four events: 1. The Berlin Wall fell 2. The Challenger disaster 3. The Chernobyl disaster 4. The Soviet Union was dissolved
The Berlin Wall fell0.00
✓ The Challenger disaster0.56
The Chernobyl disaster0.44
The Soviet Union was dissolved0.00
earliest of four✗ wrongp(correct) 0.09
Four events: 1. The euro banknotes entered circulation 2. Wikipedia launched 3. The first iPod was released 4. The September 11 attacks
The euro banknotes entered circulation0.38
✓ Wikipedia launched0.09
The first iPod was released0.02
The September 11 attacks0.51
closest pair✓ correctp(correct) 0.84
Four pairs of events: 1. ChatGPT's release and Threads' launch 2. The first iPhone and Instagram's launch 3. Wikipedia's launch and Facebook's founding 4. Sputnik 1 and Apollo 11
✓ ChatGPT's release and Threads' launch0.84
The first iPhone and Instagram's launch0.03
Wikipedia's launch and Facebook's founding0.13
Sputnik 1 and Apollo 110.00
closest pair✗ wrongp(correct) 0.14
Four pairs of events: 1. The Challenger and Chernobyl disasters 2. Facebook's founding and YouTube's founding 3. The September 11 attacks and the first iPod 4. The Berlin Wall's fall and the Soviet Union's dissolution
The Challenger and Chernobyl disasters0.55
Facebook's founding and YouTube's founding0.12
✓ The September 11 attacks and the first iPod0.14
The Berlin Wall's fall and the Soviet Union's dissolution0.19
interpolate the year✓ correctp(correct) 0.66
An unnamed product launched after the first iPhone went on sale, before Instagram launched, and around the same time as Bitcoin's first year.
20050.01
20070.32
✓ 20090.66
20110.01
20130.00
interpolate the year✓ correctp(correct) 0.69
An unnamed AI model was released after ChatGPT but before Meta's Threads app launched.
20200.00
20210.01
20220.29
✓ 20230.69
20240.01
interpolate the year✓ correctp(correct) 0.89
An unnamed games console was released after the PlayStation 2 and before the PlayStation 3.
19990.00
✓ 20030.89
20070.11
20110.00
20150.00
interpolate the year✓ correctp(correct) 0.98
An unnamed film came out the same summer as Barbie and Oppenheimer.
20190.00
20210.00
20220.02
✓ 20230.98
20240.00

Method and caveats

  • Real titles come from Wikipedia’s List of American films of 2023–2026, up to twenty per month, preferring films with more linked cast and crew. Invented titles are a fixed list of ninety; each question draws three at random.
  • Famous people come from Wikipedia’s monthly death lists (~800 a month) ranked by article length × language editions; Jev has no opinion about most of them (below ~2M the answer is a coin toss on either side of the cutoff), so the chart keeps the famous subset. A long public illness biases toward “dead” regardless of date (Brian Wilson 0.83, Ted Turner 0.87), and a bare name can’t distinguish very old from dead.
  • Recognition never reaches chance in 2026 (≈0.36): films are announced and trailered long before release, and invented titles may still read as slightly “off”. Both push the floor above 0.25.
  • The decline starts well before the midpoint: 2023 ≈ 0.81, 2024 ≈ 0.71. Coverage of a film keeps growing for years after release, so recency alone lowers recognition. The horizon is where that slope steepens, not a wall.
  • jev-preview was run too and matches jev-latest within ±0.01 everywhere; it is not shown.
  • Chronology dates are hand-curated and all ≤ 2023 so ground truth doesn’t depend on any model’s cutoff, including the one that wrote this page.