One question: could Flowers Foods (Nature's Own, Dave's Killer Bread, Wonder) double to $12 by January 2029? From 5:32pm to 1:25am a swarm of Claude and Codex agents read SEC filings, county tax rolls, trademark records and Walmart shelves, chased every lead it found, argued with itself, then red-teamed every bull case it had built.
This is how the agents workWatch 305 Claude and Codex agents research FLO overnight: the minute-by-minute shift, one lead traced from idea to verdict, a replay of an agent at work, and the war game. Inside the swarm ↓The options market already prices about 14% for $12+. To have an edge you'd need to believe 20–25% or more, and only one of seven sources got close.
Bottom line: Flowers Foods has about a 9% chance of closing at $12 or more on January 19, 2029, double today's $6.11. The options market already prices about 14%, so the double isn't underpriced. The most likely outcome is a long drift between $5 and $8.60 while you collect the dividend.
One thing could change that: Bimbo, the biggest US baker, raising bread prices in 2027. Its earnings call on Thursday, October 22 is the first test.
Flowers spent a decade buying growth while its core business shrank. It's the #2 US bread maker behind Bimbo, with Nature's Own, Dave's Killer Bread, Wonder, Canyon Bakehouse and Tastykake. From 2015 to 2025 it spent $1.66B on acquisitions, ending with about $850M for Simple Mills. Operating income plus D&A only grew from $430M to $485M.
Now the decline shows up everywhere. Organic sales have fallen nine straight quarters, and branded volume was down 7.6% last quarter. Debt is 3.5–4.0x EBITDA, the highest in at least 15 years, and all three rating agencies had cut Flowers to junk by mid-September.
Shareholders have paid for it. The dividend, raised in May 2025, was halved in May 2026. The stock is down 82% from its peak and about 23% of the float is short. Since the current CEO took over in 2019, shareholders have lost 63% while the S&P 500 returned 202%.
$6.11 doesn't price a company in permanent decline. It assumes earnings take one more step down in FY27 and then level off, with EBITDA around $413–446M. At 6.3–7.1x EBITDA, Flowers trades about where Bimbo does.
So the question isn't whether it's cheap. It's whether earnings stop falling. In past turnarounds like this, the dividend cut, the junk rating and the short interest didn't decide the outcome; the next two years of earnings did. The closest winner is Molson Coors in 2020: a similar multiple, similar debt and a suspended dividend, then +49% over two years once earnings held flat.
Flowers can't raise prices on its own. It tried in 2026. Wholesale white bread prices rose 6.84% in September, but the average retail price fell 1.98% (BLS), because Bimbo didn't follow.
Bimbo has little reason to change course. It's winning on cost, with a 10.0% North America margin and US share gains in every category last quarter. One agent also found, in Bimbo's filing with the Mexican stock exchange, that it had hedged its wheat through June 2027 before prices spiked.
That makes 10/22 the big tell. If Bimbo announces US price increases for 2027, our odds of $12 rise to about 13–15%. If it talks about "productivity" and share gains again, they fall to about 5%.
About $490–500M of EBITDA at 8x, against roughly $430–441M expected for FY27. Closing that gap takes five things going right in order, starting with Bimbo. After two rounds of red-teaming, the Bimbo route was the only one that could reach $12 on its own, about 2% of the time. The next section shows how every estimate stacks up.
Sources: SEC filings, BLS, Grupo Bimbo BMV filings, FINRA, FMP and S&P consensus via MarketScreener, as of 10/17/26. FLO closed at $6.11 on 10/16.
No single estimate was trusted on its own, so the run produced seven sets of odds for $12 or more: the agent fleet, two red teams, two pessimism checks, a war game, and the options market itself. They land between 5% and 18%. None reaches the 20–25% you would need to have a real edge over what options already price.
Price only, estimates. The $3–5 bucket is from the Claude red team and the $8.60–12 bucket from the fleet calibration, so the buckets sum to ~99% after rounding. Add roughly $0.90–1.10 of dividends to any outcome where the dividend survives.
| Route | Before | After | |
|---|---|---|---|
| Bimbo-led pricing + volume plateau | 3.5% | ~2–3% | survives |
| Network reset via activist or new CEO | 2.5% | ~1–2% | split |
| Whole-company takeout | 1.5% | ~0.5–3% | split |
| Wheat and cost relief kept | ~1% | <0.5% | killed |
| Spin or sell the premium brands | ~0.5% | ~0.4% | killed |
| $20: everything works, then a buyer | ~0.5% | ~0.3% | inside the others |
Odds of reaching $12+ by each route (est). "Split" means Claude's red team killed it and Codex's kept it. A takeout most likely lands at $8.50–9.50, so it works as a finisher for the first two routes, not on its own.
The most interesting finding wasn't in the business. It was in the pay plan. One agent read the 2026 proxy, the 10-K and every award agreement since 2024, then worked out what each executive gets under different outcomes. All of them do best if the company is sold. For the CEO, a sale is worth three to five times what staying is.
Staying pays almost nothing. All three stock awards tied to relative returns are tracking zero, the 2026 return-on-capital award is near zero, and last year's bonus paid 56% of target.
Only a sale of the whole company pays. Asset sales and spin-offs don't trigger the payout, so selling Tastykake pays the executives nothing. Five of the six top executives can also retire and keep their stock awards with six months' notice, so none of them is locked in.
It gives the board a reason to settle with an activist. A dissident who wins a board majority in a proxy fight counts as a change of control. Directors added through a settlement don't.
The catch: nobody has shown up to buy. None of the ~645 institutions filing 13Fs through Q2'26 is a private equity firm, strategic buyer or sovereign fund. A buyout fund aiming for 20% returns on flat earnings could pay only $3.84–5.07 a share, below today's price.
So a deal finishes a recovery; it can't replace one. We put the odds of any deal by January 2029 at 6–15%, most likely at $8.50–9.50. It would also have to go through the old Flowers family. Family-linked holders own about 9–12%, and together with BlackRock, Vanguard and State Street they control about 41% of the vote.
Source: FY25 10-K, 2026 proxy and award agreements (SEC EDGAR); 13F holder screen through Q2'26. As of 10/17/26.
The fit is real, and the angle is worth keeping open. Kurt runs Hestia Capital. He won 4 of 9 Pitney Bowes board seats in 2023 and has been Pitney's CEO since May 2025. The hunch is that Flowers is perfect for him and the timing is right. The agents agree on the fit. Full control is the stretch.
Flowers checks his boxes. Returns have been poor under this CEO, a decade of deals didn't grow profit, and the CEO is also chairman of a company his father ran. Kurt is buying, too: Hestia went from 31K to 363K FLO shares in the year to June 2026.
The calendar lines up. His new fund, Helios II, took in its first money on 9/29/26. At Pitney Bowes, his 13D came 49 days after the equivalent fund launched, which would put a first public move around 11/17–12/8/26. Board nominations are due 1/29–2/28/27.
Control is the hard part. Hestia owns 0.17% of Flowers, against 2.26% of Pitney at the same stage, and even all of Helios II's $9.5M only gets it to about 0.9%. Activists win about 1.1 seats on average, and four unapproved directors would trip the change-of-control clause in Flowers' bank loan.
There are other frictions. Flowers passes his own screen (five years of free cash flow above the market cap) only below about $5.75. At Pitney he had a money-losing unit to cut; Flowers' problem is its core. And in February he signed a new three-year contract as Pitney's CEO, running to 2029. A Pitney sale would free up his time.
The realistic version is a settlement, and that could matter a lot. Two operator directors pushing for bakery closures and a Simple Mills sale is exactly the catalyst the bull case needs. Recall that the pay plan already gives the board a reason to settle rather than fight.
How we'll know. First, Hestia's Q3 13F, due by 11/16: under 363K shares is a negative, 0.4–1.0M is consistent, over 1.0M is a strong sign. Then a 13D or public letter in the 11/17–12/8 window, and nominations by 2/28/27.
The risk is earnings, not the multiple. At 6.3–7.1x EBITDA there isn't much valuation left to lose, but there's plenty of earnings. Codex's red team puts the odds of a close below $5 at 32%.
The nearest test is Flowers' Q3 report in November. The real bar for FY26 EBITDA is about $460M, not the $433.7M on FMP, so a $445M guide would be a miss worth about $0.45 a share. We put the odds that the low end gets cut at 40–65%.
Flowers may already be cutting price. An H-E-B check on 10/18 found Nature's Own at $3.37, down from $3.76, with a 2-for-$5 deal. That's one store, but if it spreads to Walmart and Kroger, Flowers closes the gap with Bimbo at its own expense.
Then there's the slower stuff. There's about a 26–30% chance the dividend is suspended by 2029, and a dividend ETF (PEY) has to sell about 6.7M shares after March 19 either way. A Massachusetts lawsuit challenging Flowers' independent-distributor model finished trial on 10/16 with no verdict yet. And in calm markets, only 1 of 30 similar lows held for a year.
The debt is less of a near-term threat than it looks. The first tighter leverage test isn't until about 1/1/28, and suspending the dividend would free about $106M a year.
A few things the market may be too gloomy about. The 2031 bonds at 84.3 imply 6–10% default odds by 2029, but the loan terms have more room than bears assume. About $130M of loans to Flowers' own distributors sits outside most valuation math. Volume comparisons get easier over the next three quarters. And with 23% of the float short, any sign of stabilization could set off fast covering.
On the big picture, the market is probably right. Flowers isn't cheaper than Bimbo, the 2026 price increase already failed once, and the odds roughly match what similar setups have done. Any edge has to come from news specific to Flowers, which is why the dates below matter.
Three dates decide most of this: Bimbo on 10/22, Flowers' own Q3 in November, and the FY27 guide in February, which lands inside the window for board nominations.
Bimbo's Q3 call. Explicit US price increases for 2027 is the single biggest tell. "Productivity" and share gains point down.
Pitney Bowes Q3. The sale-review timeline tells you how much of Kurt's time is spoken for.
Hestia's Q3 13F. Under 363K FLO shares is negative, 0.4–1.0M is consistent, over 1.0M is strong confirmation.
FLO Q3 and the November guide. Keep the $453M EBITDA low end and branded volume better than −7%. A cut below ~$445M is a miss.
The Kurt window. On his last campaign's pace, a 13D or open letter would land here.
Nomination window. If the board fills its open seat with an operator before this, the odds of a proxy fight roughly halve. The real confirm or kill for activism.
FY27 guide. Midpoint $440M+ and 4 or more named closures is the bull read. Under $420M is the bear read.
Dividend ETF (PEY) drops FLO. About 6.7M shares of forced selling, even if the dividend holds.
Q1 FY27 branded net sales. 0% or better means volume has stopped falling.
Bimbo deadline. No US price move by now kills the main route to $12.
Everything above came out of one overnight run, built on two earlier deep dives. Here's how the agents worked, coordinated and argued, minute by minute.
This Mac lets one workflow run 6 agents at a time (8 CPUs), so three Claude workflows ran side by side and Codex worked on top of them. Hover or tap the chart to see who was working.
Every agent read the same brief: what an earlier 14-lane deep dive already knew, the bar to beat ($12 is about 9x the $467M EBITDA guide), and the rules. Read-only, every number carries a date and a source, and each claim gets marked verified, reported or inferred.
One creative lane each. They raised 106 leads.
Leads were clustered, then each digger returned a verdict plus new leads to chase, up to 3 levels deep. This is what blew through the usage limit.
Where the market could be wrong vs past situations, technicals, who's buying, smart money, exec incentives, odds implied by options, a Q3 earnings nowcast.
11 agents role-played the CEO, CFO, board, Bimbo, Kurt, a PE fund, a strategic buyer, Walmart and Kroger buyers, short sellers, the banks and long holders, across 5 simulations. Then 6 agents wrote histories from 2029 working backward from a $12 outcome, and a checker tested each against today's evidence.
Fit, timing, control, a red team, a verdict. Would the Pitney Bowes activist actually love this company?
GPT at its highest reasoning setting took the same brief: 15 independent reports on the big questions, then the 168-lead backlog.
Every verdict from both model families merged into six routes to a double, plus a base case and a downside, each with odds, must-be-trues and kill conditions.
A Claude red team attacked each route, a Codex red team attacked the bull case, and both families ran a pessimism check on the bears.
Reconciled seven sets of odds into one call and logged every number the run corrected.
80 verdicts from Claude diggers, 168 from Codex. Separately, the agents logged 211 dead ends. The 248 leads became 8 routes, and 1 route survived the red teams on its own.
The biggest single swing for Flowers is whether Bimbo, the biggest US baker, raises prices in 2027 so Flowers can follow. Here is that one question moving through the swarm across five and a half hours: who raised it, who chased it, who argued with it, and how the odds moved at each step.
The plant-capacity, costs-and-policy and fallen-angels scouts each came back with "Bimbo leads a 2027 price increase", each with different odds and a different value. A fourth scout, studying turnaround playbooks, raised the mirror image: Bimbo uses productivity to hold its price gap and keeps taking share, at 40–50%. The network outage killed two of them mid-work, and their retries picked up from the files the first attempts had saved.
"These three are near-duplicates. Establish whether Grupo Bimbo signals or takes a 2027 US list-price increase, and what share of FLO's ~$105–200M FY27 inflation basket can then pass through."
106 leads became 53 groups to dig, each with a brief on where to start looking.
It read Bimbo's earnings calls back to 2012, BLS price files, Bimbo's filings with the Mexican stock exchange and old Walmart pages on the Wayback Machine. Buried in the Mexican filing was a wheat futures table: Bimbo had locked in 10,543 contracts (about 52.7M bushels) for December 2026 to June 2027, all added in Q2, before wheat spiked. Flowers' own 10-Q showed just $0.4M of hedges for FY27. If Bimbo's costs are locked, it has less reason to raise prices. Watch the replay below.
"Bimbo-led 2027 pricing plus a volume plateau, re-rated." About 10% that it happens in some form, and 3.5% that it alone gets FLO to $12 or more.
It also found three mistakes upstream. Bimbo's hedge book is a rolling program, so a December 2025 snapshot already showed 11,175 contracts out to December 2026. So "Bimbo's cover runs out mid-2027" was probably wrong: the cover will likely roll forward. The value math also counted Tastykake's sale proceeds and its profit at the same time. And a peer multiple of 8.09x couldn't be reproduced; FMP shows 7.17x.
Strongest counter-evidence: "The 2026 industry list increase already failed at the shelf." Wholesale white bread +6.84% y/y, retail −1.98%.
The tell is Bimbo's 10/22 call. If it announces US price increases for 2027, this route roughly doubles and total odds of $12+ move to about 13–15%. If it only talks productivity, total odds fall to about 5%.
This is the dig agent's own transcript, condensed from 76 tool calls to 24 steps, with real timestamps and some of its actual commands. Nobody told it to look at Mexican exchange filings. It decided wheat hedges mattered, guessed where Bimbo would disclose them, and searched the filings for "trigo", Spanish for wheat.
Every agent ends with a "could not verify" list. This one's included: Bimbo's Q3 call (not out until 10/22), which exchange its wheat contracts trade on, and Flowers' off-balance-sheet flour contracts.
The agents never message each other. A short JavaScript workflow decides who runs when, and everything passes through files and structured hand-offs. That's what lets 300 of them work on one problem without stepping on each other.
{
"id": "CAP-1",
"title": "Bimbo Q3'26 call (10/22, 5pm ET): 2027 US pricing stance",
"claim": "Bimbo signals or takes US list-price increases for early 2027 (≥3%) to cover wheat, packaging and labor … FLO's FY27 inflation (~$105–200M) then passes through instead of being absorbed.",
"value_per_share_if_true": "+$1.5–3.0/sh relative to a no-follow FY27, i.e. $50–100M of EBITDA preserved at 7x (est)",
"odds": "25–35% (est). Current rhetoric leans to productivity over price.",
"evidence_so_far": "Q4'25: 'rational and disciplined…' Q2'26: 'very promotional' (analyst); 'another P… productivity' (CEO). NA margin 10.0%, still ~270bp below 2021's 12.7%.",
"next_tests": [
"Read the Grupo Bimbo 3Q26 release and transcript …",
"Pull BLS PPI PCU31181231181215 … a second January step-up equals an industry list increase",
… 2 more
]
}
Every lead carries its own value per share, odds and next tests, so the next agent knows what would prove or kill it.
Each player agent researched its character and wrote a playbook with odds on every move. Then five game-master agents each wrote their own simulation engine in Python, calibrated it on 20,000 runs, fixed a random seed before starting, and played one path through to January 2029 with no re-rolls. Four games started from a premise; one was played straight.
| Game | Premise | Jan 2029 | What happened |
|---|---|---|---|
| Sim 1 | Played straight | $8.13 | Bimbo led pricing but volume kept falling. Dividend suspended May 2027, then Tastykake ($295M) and Simple Mills ($531M) were sold. No activist, CEO kept. Never got near $12. |
| Sim 2 | Weak Q3 print | $8.46 | The stock bottomed near $4 in Q3 2027. The dividend was suspended, Hestia won 2 seats in a proxy fight and the banks amended covenants. A lucky FY28 rescued it. |
| Sim 3 | Hestia files a 13D | $8.51 | The board fought, the contest went to a vote, Hestia won 2 seats (3 by 2028) and the CEO was unseated. Bimbo pricing did most of the work. |
| Sim 4 | No activist ever appears | $6.34 | A grind. Bimbo held a price umbrella, but the internal levers were never pulled. Price about flat, +22% with dividends. |
| Sim 5 | Costs fall and Bimbo raises prices | $3.14 | The friendliest premise still crashed. Volume losses and retailer give-backs swamped the price gains. |
Each was handed an ending ("it's January 2029 and FLO is at $20") and had to write the history that got there. A checker agent then tested every condition against evidence through 10/17.
The point of creative lanes is to look where analysts don't. Most of it came back dead, which is useful too: each one is a bull idea you can stop hoping for.
Wholesale white bread prices were up 6.84% y/y in September while the retail shelf price was down 1.98%. The 2026 price hike never reached shoppers.
bearishShelf checks priced Nature's Own against Sara Lee. On 10/18 H-E-B had Nature's Own cut to $3.37 from $3.76 with a 2-for-$5 deal. Flowers is closing the gap at its own expense.
bearishFamily ties in 2017–21 proxies (two legacy holders married sisters, a director is a nephew) add up to a 9–12% family bloc. It likely matches the 8–10M-share abstention on the CEO vote. Any deal runs through them.
key findingLooked for private equity, strategic buyers, Japanese trading houses and Gulf sovereign funds holding FLO. Found none. Norway's fund exited.
no buyer yetValued 22 owned depots at ~$0.9–1.0M each. Data-center land in Phoenix and Atlanta trades at $100–460K an acre, below what Flowers' bakeries there are already worth.
dead endEvery WONDER mark is filed for bakery goods only. No hidden licensing or merch business.
dead end33 federal cases since 2024 name a Flowers entity, and Flowers is the defendant in every one. No legal windfall coming.
dead endLayoff notices from 47 state agencies show no Bimbo US plant closures in 2026. Air permits lag too badly to flag closures early.
dead endOklahoma is 95% in drought vs 29% a year ago as the 2027 wheat crop goes in. No new-crop wheat relief to count on.
bearishOf 165 live Flowers postings, none are in pricing, M&A or network strategy. The Chief Supply Chain Officer seat has been empty since 4/17/26.
watchKurt's new fund, Helios II, raised $9.5M from six investors (10/6/26). That's small for a company worth $1.3B.
watchCodex found that the first covenant test falls at FY27 year-end (~1/1/28), not 10/9/27. It also found that 4 unapproved directors, not 5, would break the board-majority clause.
less bearishThe second model family paid off because the two failed in different places. Claude was better on calibration, buyers and governance. Codex was better on operating math and reading the credit documents.
| Topic | Claude said | Codex said | Better supported |
|---|---|---|---|
| Odds of $12+ | 9–10%, options ~14% | 25% in its first pass | Claude Codex's own red and blue teams landed at 7–18% |
| Profit lost per point of volume | $25M per point of branded volume | $25M is total volume; branded is ~$17M | Codex |
| First covenant test | 10/9/27 | FY27 year-end, ~1/1/28 | Codex section 10.07 wording |
| Directors needed to trip the loan | 5 of 8 | 4 of 8 | Codex |
| Kurt's Pitney commitment | 3-year contract to 2/17/2029 | An at-will 2025 letter | Both the term is real; it's still at will |
| Takeover odds | 10–13% any deal | 20% at ~$9.50 | Both Claude on odds; Codex right that $12 needs only 7.7x on a recovered base |
| Family voting bloc | 8.9–12.5% via proxy genealogy | ~1% (CEO's family only) | Claude |
| Antitrust limit on a Bimbo deal | A hard "4-point" share cap | No such rule; the 2011 decree allowed 9–10 points | Codex |