Zeitgeist — Friday, October 23, 2026
A weekly look (new edition every Monday morning) at where the system's head is at: the mood, what's working and what isn't, the themes, and the week ahead. It's all built from TRAI's own data, then rapped over an original boom-bap beat. New to markets? Start with the plain-words version. Screening/research, not investment advice.
What's going on now and next week, in plain words
Think of the VIX as the stock market's anxiety meter. It's worked out from what investors pay for protection on the S&P 500, a basket of 500 big U.S. companies. TRAI's latest check (Oct 23, 2026, 8:53 AM ET) read 19.98, which TRAI calls 'transitional' (15 to 22): more uneasy than calm, but not panic. TRAI also keeps one 'mood score' for the market, from −100 (scared) to +100 (confident). It fell from 47.6 on Oct 21 to 17.5 on Oct 22, so the mood went from 'risk-on' to 'neutral'. Prices also aren't heading clearly up or down. TRAI calls that 'choppy'.
On Thursday, Oct 22 (the latest full trading day in this edition), the S&P 500 fund (SPY) fell 1.18%. The tech-heavy Nasdaq-100 fund (QQQ) fell 2.86%, and a fund of chip makers (SMH) fell 5.99%. Among the 156 big companies TRAI follows, the biggest drops were Moderna (−17.57%), Lam Research (−11.76%) and GE Vernova (−10.64%). The biggest gains were Spotify (+11.10%) and Blackstone (+8.38%).
Some stocks move together, like birds in a flock. TRAI measures this with 'correlation': 0 means two stocks ignore each other, 1 means they move in lockstep. Its biggest flock right now has 24 stocks with an average correlation of 0.588, mostly chip makers (17 of them), plus a few from other industries, such as Caterpillar, Arista Networks and Vertiv. Owning several of them is really one big bet, which is why TRAI allows only one trade per flock (a 'correlation cluster').
Over the last 20 trading days (about a month), investors leaned toward 'defensive' businesses, the ones people pay for in good times and bad. Compared with the S&P 500, consumer staples (everyday basics like food, drinks and soap) did 10.6 points better, utilities (power and water companies) did 9.3 points better and energy 8.4 better. Chip stocks did 12.4 points worse and technology did 5.9 points worse. Meanwhile the 10-year Treasury yield, the interest rate the U.S. government pays to borrow for 10 years, was 4.96%, 0.14 points lower than a week earlier. It matters because it nudges the cost of mortgages and business loans.
TRAI also asks how unusual a stock's move was once you take out what the whole market did. That's the 'residual z-score'. Near 0 is ordinary, 2 is rare and above 3 is very rare. Blackstone (BX) rose 8.38% with a score of 6.19, a standout on a day the market fell.
TRAI tests its trading styles on past data, like testing a recipe by cooking it many times. In the test from Apr 23 to Oct 21 (126 trading days), the weakest style was short momentum: betting against stocks that just closed weak on heavy trading. Its 412 test trades lost 0.46% each on average after trading costs and won 43% of the time, and the result is clear enough that it isn't just bad luck. The best was the 'buy the dip' fade (buying a stock the day after a big one-day drop that happened on normal trading), at +0.165% per trade, which is too small a sample to call proven. Live tracking has only just begun: 35 real-time picks are still waiting for results.
Earnings are a company's report card. Every three months it tells investors how much it sold and earned, and the stock can jump or drop on the news. 46 of the companies TRAI follows report next week (Oct 26–Nov 2). The biggest names: Microsoft and Alphabet (Google) on Wednesday after the market closes; Amazon on Thursday after the market closes; Apple on Monday Nov 2 after the market closes. Options prices (options are contracts that pay off if a stock moves a lot) suggest swings of about 7.7% for Amazon, 6.2% for Alphabet (Google) and 5.5% for Microsoft, up or down. Also: On Wednesday, the Federal Reserve (the Fed), America's central bank, announces its interest-rate decision at 2:00 PM ET, with a press conference at 2:30. On Thursday, the government reports how fast the economy grew (GDP) and how fast prices are rising (PCE inflation) at 8:30 AM ET.
TRAI's own rules work like seatbelts. Every trade has a stop loss, a pre-set exit if the price moves the wrong way by about one normal day's worth of wiggle. A trade can risk only 0.25% of the account, or $25 on $10,000. TRAI stays out of the first half hour after the 9:30 AM open, holds at most 3 trades, and closes everything by 3:55 PM, so it never holds a stock overnight when report cards come out.
Mini glossary
- VIX
- The market's anxiety meter. It's based on what investors pay for protection on the S&P 500. Higher means more nervous.
- Stock / share
- A small piece of ownership in a company. Its price goes up and down as people buy and sell it.
- S&P 500 (SPY)
- An index of 500 large U.S. companies, often used as shorthand for 'the market'. SPY is a fund that tracks it.
- Calm / transitional / stress
- TRAI's three VIX zones: calm below 15, transitional from 15 to 22, stress above 22. In stress TRAI bets half as much.
- Regime score (mood score)
- TRAI's one-number summary of market mood, from −100 to +100. +25 or more is 'risk-on' (confident), −25 or less 'risk-off' (scared), in between 'neutral'.
- Risk-on / neutral / risk-off
- Shorthand for confident, in-between and scared markets.
- Trend / choppy / ADX
- ADX measures how strongly prices are heading one way. Under 20 is 'choppy' (zig-zagging); 25 or more is 'trending'.
- ETF / fund
- A basket of many stocks you can buy in one go. SPY holds the S&P 500, QQQ the Nasdaq-100 and SMH chip makers.
- Nasdaq-100 (QQQ)
- An index of 100 big companies, heavy on technology. QQQ is a fund that tracks it.
- Semiconductors / chips (SMH)
- Companies that design or make computer chips, the brains inside phones, cars and AI data centers.
- Correlation (ρ, 'rho')
- How closely two stocks move together, from 0 (unrelated) to 1 (lockstep). 0.5 or more is a strong link.
- Correlation cluster
- A 'flock' of stocks that move together. Owning several is like one big bet, so TRAI takes at most one trade per cluster.
- Defensive stocks
- Businesses people keep paying in good times and bad, like food, soap and electricity. Investors often move into them when nervous.
- Basis points / points
- Here, 'points' means percentage points: 10 points better means a return of, say, 12% instead of 2% over the same time. (A basis point is a hundredth of one percentage point.)
- Treasury yield (10-year)
- The interest rate the U.S. government pays to borrow. The 10-year rate affects mortgages and loan costs across the economy.
- Residual z-score
- How unusual a stock's move was after removing what the whole market did. Near 0 is ordinary, 2 is rare, 3+ is very rare.
- Short / shorting
- A bet that a price will fall. You profit if it drops and lose if it rises.
- Momentum
- The idea that a stock moving strongly, on heavy trading, keeps going for a while.
- After costs
- Results after subtracting estimated trading fees and the price you lose getting in and out.
- Mean reversion / fade
- Betting that a big one-day move snaps partly back, like a stretched rubber band.
- Expectancy (per trade)
- The average gain or loss per trade. Small numbers add up over many trades, in either direction.
- Earnings (report card)
- A company's quarterly results: sales, profit and outlook. Stocks often jump or drop on them.
- Options / implied move / straddle
- Options are contracts that pay off if a stock moves. Their prices show how big a move traders expect, the 'implied move', in either direction.
- The Fed / FOMC
- The Federal Reserve, America's central bank. Its committee (the FOMC) sets short-term interest rates; markets watch every decision.
- GDP
- Gross domestic product: the total value of everything the economy produces. It measures growth.
- PCE inflation
- A government measure of how fast prices are rising, and the Fed's favorite inflation gauge.
- Stop loss
- A pre-set exit: if the price moves against you by a set amount, you get out automatically and take a small loss.
- Risk per trade
- The most a single trade is allowed to lose if its stop is hit, as a share of the account.
- Opening range / no-entry window
- The first 30 minutes after the 9:30 AM ET open, when prices jump around most. TRAI won't start trades then.
- Flat / overnight hold
- Being 'flat' means holding nothing. TRAI closes everything by 3:55 PM so nothing is held overnight.
- Long
- The normal kind of bet: buying something because you think it will rise.
- Sector
- A group of companies in the same line of business, such as energy or banks.
- Proposal (S-0xx)
- An idea from TRAI's research notes for changing the rules. It isn't a rule until it's tested and adopted.
- Position
- A trade you currently hold.
- Rotation
- When money moves from one part of the market to another instead of leaving stocks altogether.
- Backtest / replay / walk-forward
- Testing rules on past data as if trading back then, day by day, without peeking ahead. It's a test, not a promise.
- Signal / bucket
- A stock that matched one of TRAI's screening recipes on a given day. Each recipe is a 'bucket'.
- Volume / relative volume
- How many shares traded. Relative volume compares today with normal. 3× means three times the usual trading.
- t-stat
- A check on whether a result is probably real or just luck. Beyond +2 or −2 is usually taken as a real pattern.
- Hit rate
- The share of trades that made money. 50% means half won.
- Breadth / moving average
- Breadth asks how many stocks are joining in. A 20-day or 50-day average is a stock's average price over that many days; being above it is a sign of strength.
- Before open / after close (BMO / AMC)
- When a company releases earnings: before the 9:30 AM ET open, or after the 4:00 PM ET close.
- Gap / gap fill
- A gap is when a stock opens far from yesterday's close, often after news. It 'fills' if the price comes back to that close the same day.
- Mega-cap
- The very largest companies, worth hundreds of billions of dollars or more.
- ATR (average true range)
- A stock's typical daily price wiggle over the last 14 days. TRAI sets its exits in ATRs.
- Target
- A pre-set price where you take your profit and leave.
- Time stop
- If a trade hasn't worked within a set time (60 minutes for TRAI), close it anyway.
- Equity (account)
- The total value of the trading account.
- Daily loss limit
- If the account is down 1% for the day, TRAI stops trading until tomorrow.
- Term structure / contango
- Comparing short-term and longer-term VIX. Normally short-term worry is lower ('contango'). If it flips higher, people are scared right now.
- VWAP
- Volume-weighted average price: the day's average price, weighted by how much traded at each price.
- Survivorship bias
- Testing only on companies that exist today, which can make the past look better than it was.
- Universe
- The list of about 150 large, heavily traded U.S. stocks that TRAI watches.
The overview
The mood
Neutral, choppy and less confident. TRAI's composite regime score dropped from 47.6 (risk-on) on Oct 21 to 17.5 (neutral) on Oct 22. ADX is 13.5, so trend is labelled 'choppy', and breadth is split: 55.8% of the universe is above its 20-day average and 49.4% above its 50-day.
The VIX read 19.98 in the 8:53 AM run, which is the 'transitional' band. The regime model's own Oct 22 VIX input is 19.03, still transitional. Risk per trade therefore stays at the normal 0.25% (it halves only above 22). Term structure is in mild contango: VIX9D/VIX 0.944, VIX/VIX3M 0.993.
On Oct 22, SMH fell 5.99%, QQQ 2.86% and SPY 1.18%; all three closed below their 50-day averages. Over 20 days the market rotated defensively: XLP +10.6 pts vs SPY, XLU +9.3, XLE +8.4; XLK −5.9, SMH −12.4. The 10-year yield was 4.96%, −0.14 pts on the week.
What's working / what's not
Not working: short momentum. 412 replay trades averaged −0.46% each after costs (t −3.72, 43.5% winners). It was worse on transitional-VIX days (−0.51%, t −3.66) and on choppy-trend days (38.7% winners). Proposal S-002 says to skip or halve these entries when both conditions hold, as they do now.
Mildly positive, not statistically proven: the mean-reversion fade-long (+0.165%/trade, 54.3% winners, n 212, t 1.19), long momentum (+0.133%, t 0.82) and the fade-short leg (+0.097%, t 0.77). For comparison, buying everything averaged −0.01%.
Live tracking has only just started: 35 picks from the first stored run are waiting for outcomes, so there's no live track record yet. Caveats: daily-bar VWAP proxy and survivorship bias (today's universe).
The themes
Semis broke on Oct 22 and dragged the AI-power names with them. Correlation cluster 1 (24 names, average ρ 0.588) includes GEV, VRT, ANET, ETN and CAT alongside the chips. GEV (−10.64%) and VST (−9.81%) led the short-momentum bucket; LRCX fell 11.76%.
Application software (ADBE +6.07%, INTU +5.75%, NOW +2.53%) and housing/consumer names (HD +4.33%, LOW +3.97%, TGT +3.76%) rallied as yields fell. The 10:00 report found this partly unwinding by mid-morning Friday.
Friday's moves: telecom carriers fell after reports that SpaceX is seeking access to 80,000+ US cell sites, while oil services jumped on SLB and Brent above $116 (11:00 report). AXP fell after its print and dragged card peers (13:00 report). The AI-power shorts made their lows inside TRAI's 9:30–10:00 no-entry window and recovered by noon (12:00 report).
Recurring strategy themes (strategy.md, 44 entries): use correlation clusters instead of sector labels (S-001, S-023), downsize short momentum in transitional/choppy regimes (S-002), extend earnings gap-risk flags to correlated peers (S-004, S-010, S-039), don't fade big earnings gaps (S-006, S-035), and treat crowded earnings nights as one event (S-041).
The outlook
Tue 10-27: FCX, PYPL and SOFI before the open; V and HOOD after the close. Wed 10-28: GEV and VRT before the open; FOMC decision at 2:00 PM ET (press conference 2:30); GOOGL, MSFT, NOW and KLAC after the close. Thu 10-29: Q3 GDP and September PCE at 8:30 ET; MA and CAT before the open; AMZN and COIN after the close. Fri 10-30: XOM and CVX before the open. Mon 11-02: AAPL after the close. In total, 42 universe names report next week.
Options-implied earnings moves: AMZN ±7.74%, GOOGL ±6.21%, MSFT ±5.45%. TRAI's history shows GOOGL's last four earnings gaps averaged 6.37% (2.45 ATR) and none filled the same day.
What the rules say: no entries before 10:00 ET; 0.25% risk per trade (0.125% if VIX > 22); 1-ATR stop and 2-ATR target; a 60-minute time stop; at most 3 positions and 1 per correlated cluster; stop for the day at −1%; flat by 3:55 PM ET, so no position is ever held into an after-close report. The research proposals add: skip or halve short-momentum entries in today's transitional/choppy regime (S-002), don't fade big earnings gaps the same day (S-006), and on crowded nights like Wed 10-28 and Thu 10-29 allow at most one position among the reporters and their ρ ≥ 0.5 peers (S-041). Proposals are not yet engine rules.
The Zeitgeist — Eight Chambers: lyrics + plain-words translation
Original lyrics in a '90s East Coast boom-bap style. Every figure comes from the run above. Left: the bars. Right: what they mean, in plain words (open “More detail” for the full numbers).
0:00 Intro
Everything in this song comes from TRAI's automated market check of Oct 23, 2026, 8:53 AM ET. TRAI is a research tool, not a financial adviser.
More detail
Everything in this rap comes from TRAI's engine run at 8:53 AM ET on Friday, Oct 23. Its daily stats cover Thursday, Oct 22. The other inputs are the research datasets as of Oct 22–23 and today's research reports. TRAI screens stocks from public data. It doesn't recommend anything.
0:10 Verse 1 — The mood
The market's anxiety meter (VIX) is at 19.98: uneasy, not panicked. TRAI's mood score dropped from 47.6 to 17.5, from confident to undecided.
More detail
The VIX is the market's 'fear gauge', priced from S&P 500 options. TRAI's 8:53 AM run read 19.98, inside its middle 'transitional' band (15–22), so risk per trade stays at the normal 0.25%. The run's VIX9D/VIX ratio is 0.944 and VIX/VIX3M is 0.993, a mild contango, which is normal. TRAI's composite regime score blends volatility, VIX, term structure, trend and breadth on a −100…+100 scale. It fell from 47.6 (risk-on) on Oct 21 to 17.5 (neutral) on Oct 22. 'Choppy' means SPY has no strong trend (ADX 13.5; under 20 = no trend). In plain English: the market isn't panicking, but it lost its confidence in one day.
On Thursday, the chip-stock fund (SMH) fell 5.99%, Lam Research (LRCX) fell 11.76% and GE Vernova (GEV) fell 10.64%. Both are in the same 'flock' of stocks that move together.
More detail
Thursday, Oct 22 was a semiconductor selloff. SMH, the chip-stock ETF, fell 5.99%, QQQ −2.86% and SPY −1.18%. Lam Research (LRCX) lost 11.76% and GE Vernova (GEV, which makes power equipment used by AI data centers) lost 10.64%. All of these are one-day moves from TRAI's run. The 09:00 research report covers the why.
24 stocks (chip makers, industrial companies and computer-hardware makers) have been moving together like a flock. Owning several is really one big bet, so TRAI counts them as one.
More detail
TRAI groups stocks by how they actually trade together over 60 sessions. Cluster 1 has 24 names: 17 semis plus AI-hardware and power names such as GEV, VRT, ANET, ETN and CAT. Their average correlation (ρ, 'rho') is 0.588, so they tend to rise and fall together. GEV is filed under 'Industrials' but behaves like a chip stock. Strategy proposal S-001 says the 'max one position per cluster' risk cap should use these correlation clusters, not sector labels, so you don't hold the same bet twice without realising it.
Over the past month money moved into steadier businesses: everyday basics like food, drinks and soap, power and water companies and oil and gas companies beat the S&P 500. The 10-year interest rate was 4.96%, down 0.14 points on the week. Money moving between parts of the market like this is called a 'rotation'.
More detail
Over the last 20 trading days, defensive sectors beat the S&P 500: consumer staples (XLP) by 10.6 points, utilities (XLU) by 9.3 and energy (XLE) by 8.4. Tech (XLK) trailed by 5.9 and semis (SMH) by 12.4. The 10-year Treasury yield was 4.96% on Oct 22, down 0.14 points on the week. The research read is 'rotation, not liquidation': investors are moving money between sectors rather than fleeing stocks.
0:49 Hook — Eight chambers
TRAI trades by eight fixed safety rules, set ahead of time so it never makes decisions in a panic.
Don't start any trade in the first half hour after the market opens (9:30–10:00 AM ET), when prices jump around the most.
Every trade gets a stop loss (an automatic exit) about one normal day's price wiggle away, and a profit target at twice that distance, so wins are bigger than losses.
If a trade hasn't worked within 60 minutes, close it anyway.
Each trade can lose at most 0.25% of the account ($25 on $10,000), and half that when the anxiety meter (VIX) is above 22. Hold no more than 3 trades at once.
Only one trade from each 'flock' of stocks that move together. If the account is down 1% for the day, stop trading until tomorrow.
Close everything by 3:55 PM, so nothing is held overnight.
Everything here comes from public data. It's research, not financial advice.
More detail on the rules
The hook lists all eight of TRAI's risk rules from the engine run. (1) No new entries 9:30–10:00 ET; let the opening range form first. (2) Risk 0.25% of equity per trade, halved to 0.125% if the VIX is above 22. (3) Stop loss at 1× ATR (average true range, a stock's typical daily move) from entry. (4) Profit target at 2× ATR. (5) Time stop: exit if the trade isn't working after 60 minutes. (6) At most 3 positions at once, and no more than 1 per correlated cluster. (7) Down 1% on the day: stop trading. (8) Flat by 3:55 PM ET, no overnight holds. These are the 'chambers': rules that limit damage. They don't predict prices.
1:10 Verse 2 — What's working, what's not
TRAI re-ran its rules on past data, Apr 23 to Oct 21, to see which styles actually made money.
More detail
The scorecard replays TRAI's rules on history. Signals run from Apr 23 to Oct 21, 2026 (126 signal days). Each trade enters at the next session's open and exits at that session's close, net of estimated costs and market impact. Caveats: it uses daily bars only, with a stand-in for VWAP, and today's stock list, which creates survivorship bias.
The worst style was betting against stocks that just closed weak on heavy trading. It lost a little on average (0.46% a trade) and won only 43% of the time, too often to be bad luck.
More detail
Short momentum means shorting stocks that closed weak on heavy volume. It has lost money: 412 trades averaging −0.46% each after costs. The t-stat is −3.72; anything beyond ±2 is unlikely to be luck. It won only 43.5% of the time, and 38.7% in 'choppy' markets. In plain English: chasing losers lower has reliably backfired in this sample.
TRAI's research notes suggest skipping that losing style, or betting half as much, on uneasy, zig-zag days like these. It's an idea, not a rule yet.
More detail
Strategy proposal S-002, not yet an engine rule: when the VIX label is 'transitional' and the trend is 'choppy', which is exactly today's reading, skip short-momentum entries or cut their risk in half. In the replay, transitional days averaged −0.51% per short-momentum trade (t −3.66).
The best style was buying a stock the day after a big one-day drop that happened on normal trading. It made a small profit per trade (+0.165%), but there isn't enough evidence yet to call it proven.
More detail
The best-scoring style was the mean-reversion 'fade-long': buying a stock after a big one-day drop that happened on ordinary volume. It averaged +0.165% per trade with 54% winners over 212 trades. Its t-stat of 1.19 is under 2, though, so the edge isn't statistically proven. Long momentum averaged +0.133% (t 0.82).
Some stocks moved for their own reasons: Blackstone (+8.38%) and Spotify (+11.10%). Both were very unusual moves compared with the market (unusualness scores of 6.19 and 4.32; 3 or more is very rare). TRAI's 35 live picks are still waiting for results.
More detail
Today's top long-momentum name is Blackstone (BX): +8.38% on Oct 22 with a residual z-score of 6.19. Once you strip out what the overall market explains, its move was more than six standard deviations from normal, which is extremely unusual. It closed at 98.7% of its day's range on 2.12× normal volume. Spotify (SPOT) is next: residual z 4.32 on 3.23× normal volume. The live scorecard, which tracks real picks forward, has 35 picks from its first stored run still waiting for results, so there's no live track record yet.
2:10 Verse 3 — The outlook
Earnings are a company's report card. Tuesday, before the market opens: Freeport-McMoRan reports earnings. Tuesday, after the market closes: Visa and Robinhood report earnings. Wednesday, before the market opens: GE Vernova and Vertiv report earnings. Wednesday: The Federal Reserve (the Fed), America's central bank, announces its interest-rate decision at 2:00 PM ET, with a press conference at 2:30.
More detail
Next week is packed. Tue Oct 27: Freeport-McMoRan (FCX) reports before the open, and Visa (V) and Robinhood (HOOD) after the close. Wed Oct 28: GE Vernova (GEV) and Vertiv (VRT) report before the open. The Fed's two-day meeting (Oct 27–28) ends with a rate decision at 2:00 PM ET and a press conference at 2:30. All dates are from TRAI's earnings calendar and the 14:00 research report, which checked them against company announcements and the Fed calendar.
Wednesday, after the market closes: Alphabet (Google), Microsoft, ServiceNow and KLA report earnings. Thursday, after the market closes: Amazon reports earnings. Thursday: The government reports how fast the economy grew (GDP) and how fast prices are rising (PCE inflation) at 8:30 AM ET.
More detail
Wednesday after the close, Alphabet (GOOGL), Microsoft (MSFT), ServiceNow (NOW) and KLA (KLAC) all report. Thursday Oct 29 at 8:30 AM ET brings Q3 GDP and September PCE, the Fed's preferred inflation gauge. Amazon (AMZN) reports Thursday after the close. Its options price an 'implied move' of about ±7.74%: the market's estimate of how far the stock could jump, in either direction, after earnings.
Options prices say traders expect Alphabet (Google) to move about 6.21% and Microsoft to move about 5.45% after earnings, up or down. After each of Alphabet's last 4 reports, the stock never came back to its old price the same day, so betting on a quick snap-back wouldn't have worked.
More detail
Options imply about ±6.21% for Alphabet and ±5.45% for Microsoft. TRAI's earnings history shows none of Alphabet's last four earnings gaps 'filled', meaning the price didn't return to the prior close that day. Proposal S-006: don't bet against (fade) big earnings gaps on the same day.
Monday Nov 2, after the market closes: Apple reports earnings. A research idea: when several giants report the same night, treat them as one big event and hold at most one trade among them. TRAI's rules don't forecast anything; they're there to limit damage.
More detail
Apple (AAPL) reports Mon Nov 2 after the close. Proposal S-041, the 'crowded-night rule', applies when two or more mega-caps report the same night, or a reaction day lands on Fed or GDP news. In that case, treat it all as one event: at most one position among the reporters and their closely correlated peers, and no long-one/short-the-other pairs, because same-night reporters often move in opposite directions. TRAI's rules don't forecast anything. They limit how much can go wrong.
3:00 Outro
These numbers are a snapshot from Oct 23, 2026, 8:53 AM ET. This is research, not advice.
More detail
All numbers are as of the 8:53 AM ET run on Oct 23 and the research datasets it cites. The post-close run will refresh them, and this page shows when newer data exists. This is research, not advice.
Archive
- Monday, November 2, 2026
- Friday, October 23, 2026
How this is made
A script on the research box (scripts/zeitgeist/) pulls TRAI's public JSON (/api/latest, regime, scorecard, events, correlations, macro, options). It then checks every number in the lyrics and overview against that data, and refuses to render if anything has changed. The beat (drums, dusty piano, strings, bass, vinyl crackle) is synthesized from scratch in code at 88 BPM, with no samples. A neural text-to-speech voice performs the lyrics, with tickers spelled letter by letter. Each line is aligned with speech recognition and placed on the downbeat of its bar. Slides are drawn with matplotlib from the same data, and ffmpeg encodes the H.264/AAC MP4. The lyrics are original: they don't quote any existing song or imitate any real artist. Macro dates (FOMC, GDP/PCE) come from the day's research reports. The plain-words story, glossary and translations are generated by the same script from the same data on every run, with no hand-typed numbers. A scheduled run rebuilds the page every Monday morning and frames it as the week ahead. Screening/research, not investment advice.