Fit

The answer

Does not fit the framework (P1 not met); contested: P4c, P5.

AT&T clears the universe and trips no exclusion, but the year-10 durability gate (P1) is the pure binary in this system, and the jury returned it not met: revenue-higher-in-a-decade is a high-conviction call, but adjusted free cash flow materially higher with very high conviction is not, and any proper doubt fails the gate. That gate is decisive — nothing downstream offsets it. Confidence is low: the name-masked re-run flipped P1 to met and the load-bearing probability diverged past 0.20, so the run also carries a prior-driven-risk flag. Two criteria are contested (P4c, P5).

Universe and exclusions — the clean screen

Nothing here disqualifies AT&T, and no softening is needed because there is nothing to soften. Both universe lines pass and all five exclusion screens come back clean; the honest counter-facts sit inside each check rather than hidden from it.

AT&T is a Delaware-incorporated US holding company whose common stock lists on the NYSE under ticker "T" — a domestic issuer, not a Chinese company or ADR [1][2]. Market capitalization is about $164.8B — 7.179B shares at the 2026-07-23 close of $22.96 — roughly sixteen times the framework's $10B floor. The universe verdict is met on both lines.

The exclusions do not fire. The auto value-trap screen is inapplicable — this is a telecom carrier, not a vehicle maker. The darling screen passes: at ~1.3x sales ($164.8B / $125.6B) and ~7.6x earnings AT&T is a low-multiple income name, not a bottom-left-to-top-right story [3]. China dependence is absent — revenue is earned in the US and Mexico (~3% of segment revenue), with no customer above 10%. The promotional-CEO screen passes on promise-versus-delivery: four of five 2023–24 commitments were kept (fiber 32M vs a 30M+ target, 2025 FCF $16.6B vs $16B+, leverage to the 2.53x target) — though the counter-fact belongs on the record: the prior-management DirecTV (~$49B) and Time Warner (~$85B) deals destroyed value, raised the share count, and forced the 2022 dividend cut [4]. The structural-decline screen is the closest call: legacy copper is in genuine secular decline and Business Wireline revenue fell 8.4% to $17.2B in 2025 and now runs an operating loss — but that pocket is under 14% of consolidated revenue and is being deliberately decommissioned, and the three-consecutive-year decline disqualifier reads two decline years, not three [5][6].

Pattern match

Of the framework's four recognition lenses, AT&T comes nearest to the high dividend yield plus high FCF yield, business not going away setup: a 4.8% dividend and a ~10% reported FCF yield on an essential, capital-heavy oligopoly. It is not the healthcare/insurance forecasting-error pattern, and that distinction cuts against the setup rather than for it: there was no short-term earnings cut for the market to anchor to. Over the drawdown, forward EPS estimates rose (FY2027 +4.6%, FY2028 +12.8%) and AT&T beat consensus every quarter, so what repriced was terminal value, not near-term earning power [7]. It is not a quality-tech-monopoly dip and not a large-bank cyclical. The pattern it best resembles is the one whose whole case turns on the yield — and the yield is exactly where the framework's own arithmetic falls short (see the pillar ledger and Yield).

The pillar ledger

The tally's per-criterion verdicts, the reference line each sits against, and the deciding arithmetic. Reference lines, never grades.

Source: ruchir/fit_tally.json and the surviving tab claims; figures cited in the treatments below.

Year-10 durability gate (P1) — the decisive result

Here is the decisive point. The gate is binary by construction, and the jury did not clear it. Revenue-higher-in-a-decade is a very-high-conviction call: AT&T sits inside a three-carrier, FCC-licensed facilities-based oligopoly on a ~$260B network-and-spectrum base — net PP&E $131.6B plus net licenses $128.1B — that a competitor cannot cheaply replicate, with postpaid phone churn under 1% and a base that grew three straight years [8][9][10].

What the gate does not clear is the cash leg. The strongest surviving counter-fact, in the same breath: reported free cash flow has been stable but flat — roughly $14–18B every year since 2022, not rising — while capex ran 16.6% of revenue ($20.8B), net debt is ~$118B, a ~$23B EchoStar spectrum purchase is pending, and cable's two leaders added ~3.4M wireless lines in 2025 against AT&T's 1.47M, pressing the wireless ARPU that funds the enterprise [11][12]. "Adjusted FCF materially higher, with very high conviction" is the phrase under real strain, and under this gate strain is enough: three of four jurors returned not met (one met), trimmed-mean probability 0.675 with a 0.14 spread. The name-masked re-run flipped to met — that divergence is why the run flags prior_driven_risk and why confidence is low. The full treatment is in Durability. P1 not met sends the overall verdict to does not fit; no other pillar offsets a failed gate.

Consistency (P2) — met

Reported FCF held a $14.1–17.6B band with no negative years since the 2022 WarnerMedia exit, so the cash is predictable, not merely large [13]. The counter-fact carried in the same treatment: the series has definitional seams — 2024's $17.6B included ~$2.3B of since-excluded DirecTV distributions, so like-for-like 2024 was $15.3B and 2025's $16.6B is the growth — and fit_features.fcf_stability is empty with adjusted_fcf.latest_adjusted null, so P2 is evidenced on reported, not adjusted, FCF (derived: fit_features.fcf_stability). Detail in Yield.

Dislocation and fear gauge (P3a, P3b) — met, with a soft edge

A genuine dislocation exists: AT&T fell 30.9% peak-to-trough, from $29.62 (2025-09-15) to $20.48 (2026-07-01), now $22.96 (derived: fit_features.capitulation_gauge.drawdown). P3a is met. But the counter-fact is structural: the only company-documented trigger is a merely in-line Q3 FY2025 quarter that began leg 1; the far deeper leg 2 rests on press and sell-side drivers (a satellite-competition scare, an Oppenheimer downgrade, Russell index exclusion, a CFO departure), not a filed adverse event. The fear gauge (P3b) is met but muted: the peak volume spike measured 2.37x the pre-fall median, short of AT&T's own 4–5x historic panic days, though turnover did surge into the trough (~85M shares/day in July vs a ~33M pre-peak median) — mechanical selling at the low more than violent capitulation (derived: fit_features.capitulation_gauge.volume_spike). Full anatomy in Dislocation.

Adjusted yield and reversion (P3c, P3d) — the yield fails, the path is only probable

On the framework's only sanctioned basis — adjusted FCF = reported FCF − stock-based comp − the 5-year average of acquisitions — AT&T yields 5.0% against the 10% moderate bar, roughly 500 basis points short, about half the bar. The reported 10.1% ($16.6B / $164.8B) clears the bar only by ignoring the spectrum spend, and the entire gap is that spend: SBC is trivial (~$0.5B), while the trailing five-year acquisition average (~$7.9B) is ~90% the one-time 2021–22 C-band super-cycle [14][15][16]. Adjusted FCF = 16,600 − 536 − 7,871 = $8,193M; 8,193 / 164,830 = 5.0% (derived: fit_features.adjusted_fcf; fit_features.market_cap). This is the balance-sheet-appropriate bar: at 2.53x net debt to adjusted EBITDA AT&T is a moderate name, so the 10% line applies — not the fortress ~8–9% and not the levered 25% [17].

The reversion path (P3d) is met but only probable: consensus and management see reported FCF growing to ~$22B by 2028–29, carrying the framework-adjusted yield across 10% around 2028 (10.1%), which the trimmed-mean probability puts at 0.57 (spread 0.02) — the tightest agreement in the run [18]. The counter-fact: the crossing requires spectrum/M&A not to re-accelerate beyond the ~$23B EchoStar close, and a fresh C-band-scale auction would reset the trailing average and push the crossing past 2029. Workings in Yield.

Balance sheet and self-help (P4a, P4b, P4c) — outlast met, repurchases met, dividend contested

AT&T can outlast the problem without capital allocation being forced to debt paydown (P4a met): net debt ~$117.9B is 2.53x adjusted EBITDA against a 3.75x covenant, no pre-2031 maturity year exceeds ~$9B against ~$40B of operating cash flow, and $18.2B cash sits alongside undrawn facilities [19][20]. The counter-fact: the EchoStar close lifts leverage to ~3.2x before returning to the 2.5x range over ~three years, and refinancing 4.2% paper into 5.0–5.2% paper is a slow drag on free cash flow [21][22].

The repurchase engine clears the framework's hard-fail (P4b met): the share count is falling, not rising — 7.59B (2022) to 7.18B (2025), a −0.8% five-year CAGR, rising=false — and buybacks are now executed, $4.5B in 2025 and ~$10B planned for 2026 [23](derived: fit_features.share_count_trend). The counter-fact: the engine is only two years old, 2025 repurchases were done at ~$27.0/share (above today's $22.96), and the count rose ~23% from 2016–2022 on serial M&A.

P4c (dividend safety) is contested. The jury split evenly (all four jurors reading it contested). The two readings: the 4.8% dividend is well covered — $8.18B paid in 2025 against $16.6B FCF is ~2.0x cover — and management commits to hold it through 2028; against that, AT&T rebased the payout 47% (from $2.08 to $1.11) in 2022, so its record through stress is a recent cut then a hold, and 2026's ~$18B of dividends-plus-buybacks is essentially 100% of the FCF guide, leaving little room if FCF slips [24][25][26]. The report carries both readings without picking one. Detail in Self-Help.

Diagnosis (P5) — contested

P5 is contested. The temporary-or-permanent question is decided by the profile's blind adversarial trial, and the trial returned a contested ruling: three judges put the probability the impairment is temporary at 0.64, with per-judge readings of 0.64, 0.43 and 0.68 (spread 0.25) and a reading-order footprint (temporary-first 0.64 vs permanent-first 0.555 mean, gap 0.085). The temporary reading is more likely but far from settled — the permanent terminal-value reset (a ~$23–24B/yr reinvestment burden, EchoStar leverage to ~3x, and management's own two markdowns of the Business Wireline profit pool — $24.8B in 2022 and the full $4.4B of Business Wireline goodwill in 2024 on "lower long-term projected future cash flows") commands meaningful weight [27]. The Damage Math tab presents both cases and reports the 0.64 ruling without overriding it.

Instrument context (I1) — exists

Qualifying long-dated options exist: the expiration ladder runs to LEAPS dated January 21, 2028 (~18 months out), liquidity is high, and 30-day implied volatility was ~28% as of 2026-07-23 — inside the framework's up-to-~50–55 acceptable band and nowhere near 60–70. By the framework's own logic the name is expressible on the book rather than routed to a watchlist for want of instruments. The counter-fact: these facts rest on dated third-party web sources (AlphaQuery IV, a Nasdaq LEAPS article, Yahoo/Barchart chains), not a filed document, and IV can move quickly around the EchoStar close. Facts-only treatment in Clock.

What a 3x-in-3-years would require

The framework's target test — the price at the bar-yield on normalized adjusted FCF, and what consensus would have to concede — cannot be rendered here as the tally intends: re-rating math unavailable because the applicable bar or normalized adjusted FCF is missing. The adjusted-FCF series the computation needs is not_computable in this run (SBC missing for FY2016–2017; no complete consecutive five-year acquisition window), so the tally records no bar, no normalized_adjusted_fcf, and no implied market cap at the bar.

What is computable frames the same test at its edges. On the framework's adjusted basis the yield sits at 5.0% against the 10% bar — the gap the price would have to close is roughly a doubling of the adjusted yield, which the run's own path (P3d) reaches only around 2028 and only at ~57% probability, contingent on spectrum spend normalizing. And the base rate from the name's own history (Clock): the three completed same-depth drawdowns since the 2021 top recovered their pre-drawdown high on a total-return basis in 1.1 to 3.2 years from the trough — but two earlier 30%+ peaks (July 2016 −37.1%, November 2019 −32.4%) have never been reclaimed, which is the deserved-dislocation cautionary case.

Source: derived from daily price history 1990–2026, dividend/spin-adjusted (derived: fit_features.capitulation_gauge); the two open episodes remain unreclaimed as of July 2026.

Contested and undetermined

Two criteria are contested; nothing was returned cannot-determine.

• P4c (dividend safety) — the two readings above: 2.0x FCF cover and a stated hold-through-2028 commitment, against a 47% rebase in 2022 and a 2026 return plan near 100% of the FCF guide. Jury split: all four jurors contested.
• P5 (diagnosis) — temporary vs permanent, ruled at p_temporary 0.64 but contested (per-judge 0.64 / 0.43 / 0.68, spread 0.25; order-stability gap 0.085). Both readings are carried; the permanent reset holds real weight.

Provenance

Source: ruchir/fit_tally.json (provenance, mask_divergence, skeptic_counts) and ruchir/trial/tally.json.

Two model families sat the jury; a fifth, name-masked run re-scored the same evidence to catch verdicts driven by the ticker rather than the facts. Here that probe fired on the one criterion that matters most — the P1 gate flipped from not met (named) to met (masked) — so the run reports low confidence and a prior-driven-risk flag, meaning the gate verdict was pressed hard and did not settle cleanly. The skeptic recomputed the sixteen verdict-critical claims from their cited figures; all sixteen survived, none were refuted.

The falsifier ledger

The standing what-would-change-this conditions. The framework seeds, then the name-specific thresholds and windows from the run.

Framework seeds:

• adjusted FCF or EBITDA declines where flat-or-better was underwritten
• revenue declines for a third consecutive year
• capital allocation pivots to debt paydown over repurchases
• share count inflects upward
• the industry repricing cycle fails to materialize where industry-wide mean reversion was underwritten

Name-specific:

• Q3/Q4 2026 guidance cut: FY FCF below $18B or adjusted EPS below $2.25 — real consolidated earning-power loss, not a mix effect.
• Postpaid phone net adds turn negative or churn rises materially over 2-3 quarters, reversing the Q2 2026 gain and confirming share loss.
• A NEW impairment tied to lower long-term CONNECTIVITY (not just legacy wireline) cash flows, or leverage stuck above 3x post-EchoStar forcing FCF below guide.
• Wireless service-revenue growth decelerates below ~1% or ARPU declines, breaking the pricing-power basis of EBITDA growth.
• FY2027-FY2028 results deliver at least $19B/$21B of FCF with net debt/adjusted EBITDA returning toward 2.5x after EchoStar while capital investment stays within the $23-$24B range.
• Legacy revenue declines slow to low single digits and Legacy EBITDA margin stabilizes above 40% by Q4 2027.
• Advanced Connectivity sustains 5%+ service revenue growth and 6%+ EBITDA growth while postpaid phone net adds remain positive and churn does not rise materially.
• No further goodwill, license, or network-asset impairment is recorded through FY2028 for lower long-term connectivity cash flows.
• Q3/Q4 2026 guidance cut: FY FCF below $18B or adjusted EPS below $2.25.
• Mobility postpaid phone net adds turn negative or churn rises materially — impairment spreading from legacy into the dominant wireless engine.
• Consolidated adjusted EBITDA turns negative YoY, or a new goodwill/license impairment tied to lower connectivity cash flows lands through FY2028.
• EchoStar + $23-24B capex keep net-debt/EBITDA above 3x and block FCF from reaching ~$21B by 2028, confirming reinvestment permanently caps per-share NPV.

Data gaps

What the run could not answer, from the tally's list:

• fit_features.adjusted_fcf / adjusted_fcf_yield / yield_baseline / float_retirement_years are not_computable — SBC is missing for FY2016–2017 and there is no complete consecutive five-year acquisition window, so the framework's adjusted "real" FCF (the canonical yield basis, and the input the re-rating math needs) was rebuilt from filed cash-flow statements and the SBC note rather than taken from the feed.
• fit_features.revenue_trajectory is not_computable (income.json carries no annual revenue line); the disqualifier check was rebuilt from the filed income statements, yielding two consecutive decline years (2021, 2022), not the feature's stated zero.
• fit_features.balance_sheet_class is not_computable (FY2025 debt/cash absent from the feed); net-debt class was computed from the FY2025 10-K balance sheet — total debt $136.1B less cash $18.2B = ~$117.9B net, ~2.6x reported / 2.53x on management-adjusted EBITDA, i.e. moderate.
• fit_features.fcf_stability is empty (fewer than five consecutive adjusted-FCF years), so P2 rests on management-reported FCF.
• AT&T discloses no non-GAAP Free Cash Flow or consolidated Adjusted EBITDA line in its 10-Ks; reported FCF is taken from earnings releases/transcripts and adjusted EBITDA is implied from management's stated 2.53x leverage.
• Consensus positioning (X4) relied on a single web-search snapshot because the Parallel web-research pipeline failed on insufficient credit; the corpus-based multiples (7.6x P/E, 1.3x P/sales) are the primary, filing-anchored evidence.
• The CapIQ consensus-revision series spans only the trailing 180 days (from 2026-01-24, after the September 2025 peak), so estimate movement across the earliest leg of the drawdown is not directly captured.
• FINRA returned no reported short-interest rows for T in this run, so the short side of the seller composition cannot be quantified.
• Instrument facts (I1) rest on dated third-party web sources, not a filed primary document; the options-liquidity figure is a single-session snapshot and implied volatility can move quickly around the EchoStar close.
• No open-market insider purchase data to assess conviction — all recent Form 4 activity is equity grants/awards (code A).