FEBRUARY 1950

The manuscript reached the Monthly Weather Review on February 10, 1950. Three pages. A Weather Bureau author in Washington. A title that sounds like paperwork: “Verification of Forecasts Expressed in Terms of Probability.”

Glenn W. Brier had put his finger on a familiar human weakness.

Give someone a scoreboard and they will study the scoreboard. Sometimes they will study it harder than the thing they are supposed to understand.

Weather forecasters could find themselves “hedging” or “playing the system.” Brier described the temptation with a sentence that still belongs above a laboratory door:

“it is often easier to analyze the effect of different possible forecasts on the verification score than it is to analyze the weather situation.”

There it is. The weather outside. The incentives inside. Two different problems, competing for the forecaster's attention.

Brier wanted the incentives to line up. His criterion was that “the verification scheme should influence the forecaster in no undesirable way.”

So he asked the forecaster for probabilities. Rain: 70 percent. No rain: 30 percent. Then he waited for an observation. He squared the gap between each forecast probability and what actually happened, added the gaps, and averaged them across forecasts.

His table contains ten actual rain forecasts. Three occasions brought rain. Seven did not. The resulting score is 0.19, on his original scale from zero to two. Zero means perfect forecasts. Two means being absolutely certain of the wrong outcome every time.

A tiny formula. A substantial change in the bargain.

The best long-run strategy was to give the probabilities you genuinely believed. Inflating them for theatrical certainty cost you. Hiding useful knowledge behind vague numbers cost you too.

The paper appears in the January 1950 issue, which was issued on April 15. February is the manuscript-received date. The calendar matters because the lesson has survived much longer than the tools on that forecaster's desk.

Now move the desk forward to Friday, November 6, 2026.

Nace AI publishes a technical report about Drex, a family of four open-weight decision models. A decision model takes a context, a question, and a list of permitted choices, then returns a probability for each choice. The output stays inside that list.

Drex Max, the largest dense model in the family, is described as a 27-billion-parameter model. Parameters are the adjustable numbers inside a network. Here they support a system built to choose without generating an explanation word by word.

“But the number that made me stop was this one: when Drex Max says it is at least 93% sure, the company reports it is right 97% of the time.”

That is a claim worth measuring. It is also exactly the kind of claim a Weather Bureau statistician gave us a way to examine in 1950.

A WORD ON SOURCES

The historical passages and the forecast table come from the full three-page Brier paper, including a visual reading of Table 1. The modern claims come from Nace's November 6 research post, its Drex technical report, the official Drex Max model card on Hugging Face, and the release repository. Drex performance numbers below are reported by Nace unless explicitly described otherwise; the Artificial Analysis result is quoted through Nace's report of that outside evaluation and was not independently reproduced here.

A fresh X search covered November 5 through the morning of November 7, Pacific Time. It found release discussion, reposts, and questions about evaluation. The reply quoted later was returned directly by X. Quora and LinkedIn were not searched. The Rain Log script below was actually run against Brier's table and a clearly synthetic decision log. Drex itself was not run for this article.

PART ONE: THE LITTLE SQUARE THAT MAKES HONESTY PAY

The Brier score is an average squared probability error. Smaller…

In February 1950 a Weather Bureau statistician named Glenn W. Brier sent the Monthly Weather Review a three-page paper with a dry title, "Verification of Forecasts Expressed in Terms of Probability." Brier had noticed that forecasters graded by a scoreboard were tempted to "hedge" or "play the system," because "it is often easier to analyze the effect of different possible forecasts on the verification score than it is to analyze the weather situation." His fix was to ask for probabilities, wait for the weather, square the gap between each probability and what happened, and average. Brian Roemmele uses that small formula to judge a new kind of AI: Drex, a family of open-weight "decision models" that Nace AI described in a technical report on November 6, 2026.

How the score works. A forecast of 80 percent rain scores 0.04 if it rains and 0.64 if it stays dry, so confident mistakes cost the most. Lower is better. The common yes-only version runs from 0 to 1; Brier's original adds the squared errors for every outcome and runs from 0 to 2, exactly double in a yes/no problem. The key property is that honesty pays. If the true chance is 80 percent, saying 80 has an expected score of 0.16, while saying 60 or 100 both score 0.20. Over time you cannot beat the rule by stating anything other than what you really believe. Roemmele warns that the math cannot stop a dishonest record keeper who cherry-picks questions, hides forecasts or moves deadlines.

Brier's own ten forecasts. In Brier's Table 1, rain came on three of ten occasions. Two misses dominate: a 70 percent call on a dry day and a 40 percent call on a wet one. The yes-only score is 0.095 and the original score is 0.19, which matches the paper. Always forecasting 30 percent would have scored 0.42, so the forecaster was clearly adding skill. Roemmele separates three ideas: accuracy (how often the top pick wins), calibration (whether 80 percent calls come true about 80 percent of the time across many cases) and the Brier score, which rewards both.

What Drex is. Each Drex model takes a context, a question and a fixed list of allowed options, and returns one probability per option. It cannot invent an answer that is not on the list, though it can still pick the wrong one. The family is Drex DLM (8 billion parameters), v1.5 (9B), v2.0 (35B total with 3B active, a mixture-of-experts design) and Drex Max (27B). Max is built on Qwen3.8-27B, can read images and uses a special readout head, so it will not score correctly in an ordinary chat pipeline. Roemmele advises adding a "review required" option and a separate policy layer, because choosing a button is not the same as having permission to press it.

The numbers, all reported by Nace and not re-run by Roemmele. Max scores 67.52 on the public Decision Index 0.4.1, against 60.70 for Jev 1.14.0, and 66.60 with private tests added. Nace cites an Artificial Analysis score of about 64 (63.7). On Decision Bench v4 with raw images it gets 94.0 percent (1,007 of 1,071), statistically tied with d3-turbo and GPT-6 Sol and 1.6 points behind Claude Opus 5.5. The headline speed of 183 milliseconds is a separate launch measurement; on Decision Bench v4 the median is 0.66 seconds, against 2.05 for GPT-6 Sol and 4.33 for Opus. Using the actual image instead of a text description raises image-question accuracy from 63.3 to 98.3 percent (38 versus 59 of 60). Shuffling the option order changes the top answer 4.3 percent of the time, and reversing it does so 6.2 percent of the time.

The 93 percent claim. Nace says that when Max is at least 93 percent sure, it is right 97 percent of the time. Roemmele calls this a starting point. Overall, Max is mildly overconfident: 87.9 percent accuracy against an average top probability of 91.2 percent, and the smaller Drex v2.0 is better calibrated. Keeping the most confident 80 percent of answers lifts accuracy to 96.1 percent, and keeping the top half lifts it to 99.3 percent. He also flags a trap: the API's "confidence" field is a chance-adjusted transform, not the raw probability.

The Rain Log. Roemmele publishes a small standard-library Python script that reads a CSV of id, p_yes and outcome, prints both Brier conventions and a ten-bucket reliability table. He ran it on Brier's data and got 0.095 and 0.19. A synthetic, perfectly calibrated 200-row set still scores 0.1675, because calibration does not make uncertain events certain. He also offers a "Brier's Desk" prompt that makes any chat model log scoreable forecasts and write "Calibration unverified" when it has no history. His deployment advice: use separate data for building, threshold-picking and testing; track coverage and missing outcomes; and require human approval for irreversible actions whatever the confidence.

Licences, hardware and caveats. Max, v2.0 and v1.5 use Open RAIL-M terms, DLM is non-commercial CC BY-NC 4.0 and the code is Apache-2.0. A Q4 text-only build of Max runs on an Apple silicon Mac with 24 GB or more, and Q8_0 needs 48 GB. Roemmele did not run Drex itself. His business ideas (calibration audits, ticket and claims triage, local models for privacy-bound firms, scored-forecaster newsletters and decision-log tooling) are explicitly not financial advice. He closes on Brier: a rain gauge does not change the weather, but it gives an honest reading, and a decision log can do the same for an AI system.

Incoming House Judiciary chair Jamie Raskin (D-MD) said Democrats will fight to “bring government back from the brink of authoritarianism,” promising to expose what he called Trump’s pay-to-play corruption and renew government for the common good. He told MS NOW he intends to subpoena law firms that struck deals with Trump, investigate Attorney General Todd Blanche over Epstein Files Transparency Act document refusals and prosecutions of Trump’s perceived enemies (including Comey, Letitia James, Schiff, Mark Kelly, Maggie Goodlander, and Jason Crow), and work with Democratic appropriators to defund those efforts—“We have the power of the purse.”

House Oversight chair-in-waiting Robert Garcia (D-CA) listed four targets: the “Trump crime family” (citing a Russian oligarch’s payment for Don Jr.’s wedding party, Saudi investment in Kushner’s firm, Eric Trump’s Pentagon-linked board seats, and Trump’s crypto windfalls); DHS/ICE abuses; corporate corruption; and Jeffrey Epstein’s network. Sen. Ron Wyden’s Finance staff, CNBC reported, plans to dig into JPMorgan’s Epstein relationship and senior executives’ handling of it. The Senate posted a job for an attorney to lead Homeland Security ethics and anti-corruption work. CBS reported Democrats will probe Trump library fundraising: Trump once said he doesn’t believe in libraries and floated a hotel, yet his foundation (run by Eric Trump and Michael Boulos) told the IRS it is raising $950 million by end-2027, including from companies that settled suits with promises to contribute.

Trump has stayed largely quiet on the election while pushing pet projects. His hand-picked commission approved a triumphal arch so work could start immediately; Judge Tanya Chutkan then blocked construction for fourteen days. In his first post-election action he issued an order designating the Oval Office and Cabinet Room national/federal landmarks to freeze his gold-heavy renovations—changes Barack Obama once joked looked like “Saddam Hussein opened a Cheesecake Factory.”

On the right, some blame the midterm wipe-out on the Iran war or on Vance’s June memorandum of understanding. Pre-election polls, Richardson notes, put corruption at the top of complaints about the administration. Trump Media and his meme coin plunged after the results; Bloomberg’s Annie Massa said the drops suggest anxiety that Trump-linked businesses now face fresh oversight. Harmeet Dhillon resigned from Justice, where she had remade the Civil Rights Division around Trump’s social agenda, stopped defending voting rights, and pushed state voter-roll takeovers and mid-decade redistricting; more than half the division’s lawyers left on her watch.

Rep. Thomas Massie (R-KY) said he expects the signatures for a discharge petition on Epstein Files Transparency Act II when Congress returns, and vowed to force a Hegseth impeachment vote unless the defense secretary resigns first; multiple Republican sources told CNN they think the measure could pass. Hegseth posted a video calling the articles a “sham.”

Michelle Goldberg argued Elon Musk took as big a hit as Trump: nearly half a billion dollars into races, including Paxton’s Texas Senate campaign, which James Talarico won by 12 points while casting Musk as wanting the state’s land, water, and tax dollars. Markets punished the data-center build-out voters rejected—Nvidia alone lost about $400 billion (~9%); SpaceX and Tesla together more than $300 billion; the “magnificent seven” about a trillion; Nasdaq nearly $2.5 trillion.

Montana voters, meanwhile, approved a ballot measure denying corporations and other “artificial persons” the power to donate to political causes in the state—the first such measure passed directly by voters, with more than 65% yes. The Brennan Center’s Eric Petry called the near-supermajority in a red state evidence of deep public anger at money in politics.

Heather Cox Richardson opens her November 6 letter with a blunt reading of Tuesday's midterms: Americans rejected the Trump administration. In her telling, voters turned away from a strongman who ignored the Constitution, undermined democracy and enriched himself and his allies, and they did it by a margin of more than 12 million votes — demanding an end to corruption, to the war on Iran, to costly tariffs and to masked immigration agents seizing people off the streets.

She argues the numbers speak for themselves. Democrats won the national vote by more than 13 points, which she frames as a vote for the rule of law, for the Framers' checks and balances, and for policies that make food, housing and health care affordable rather than steering money to billionaires. The Economist, she notes, called it "the biggest congressional landslide in 50 years." By contrast, Trump's 2024 win over Kamala Harris — about 2.3 million votes and under 50% of the total — now looks like a fizzle next to the "mandate" he claimed.

Richardson lingers on who won. Voters elected Black Democrats Keisha Lance Bottoms in Georgia, David Crowley in Wisconsin, Aaron Ford in Nevada and Stacey Plaskett in the U.S. Virgin Islands, along with Black Republican Byron Donalds in Florida. Michigan's Abdul El-Sayed becomes the nation's first Muslim senator, and Rep. Ilhan Omar — a frequent Trump target — was reelected with 76.5% of her district's vote.

The coalition shifted too. Citing the Associated Press, she reports that Hispanic voters swung hard back toward Democrats, with seven in ten Hispanic Texans backing Senate candidate James Talarico. For the first time since 2008, Democrats won every House district touching the U.S.-Mexico border, and fourteen new Hispanic Democrats are headed to Congress. Black voters, especially in the South, turned out in unusually high numbers; the New York Times found about 80% of Black votes in House races went to Democrats. Since the election, she says, panicked Republicans have been talking about regrouping, ending the Iran war and focusing on affordability.

Trump, however, went the other direction. After appearing dazed by the result, he fired chief of staff Susie Wiles on Friday and replaced her with Stephen Miller. Richardson describes Miller as a white nationalist who appears to embrace "great replacement" thinking, the driver of the first-term Muslim ban and the second-term mass-deportation push, including the 3,000-arrests-a-day target she links to the deaths of U.S. citizens Renee Goode and Alex Pretti. She says Miller championed rule by emergency declaration, urged invoking the Insurrection Act, and — per NBC's Mike Hixenbaugh — pushed a plan to institutionalize mentally ill homeless people that led the Justice Department to stop enforcing the 1999 Olmstead disability-rights decision and drop at least five related cases, a signal to states facing Medicaid cuts. Citing Phillips O'Brien, she adds that Miller sides with Russia on Ukraine, and that his promotion suggests Trump wants to deliver what he can for Putin before the new Congress begins investigating.

She lists the circle Trump is tightening around himself: Defense Secretary Pete Hegseth, aide Natalie Harp and budget director Russell Vought, a Project 2025 author. Democrats reacted sharply — Rep. LaMonica McIver called Miller "a white supremacist," Chuck Schumer said Trump "learned nothing from his historic defeat," and Hakeem Jeffries said Miller "has American blood on his hands." One House Republican, told of the pick by Politico, asked if the reporter was joking.

Richardson reads the move as a warning to Republicans as much as Democrats: Axios reports Trump plans to keep fundraising to control 2028 candidates, and the Wall Street Journal reports the White House told more than 3,000 conservative groups it is "monitoring" their statements. "It all feels very desperate," she writes. She closes with Trump at a NASA event Friday night, in a campaign hat, rambling about nuclear spacecraft to Mars, a return to the Moon and his MIT uncle, then telling a silent room that he wasn't on the ballot — "because if we were on the ballot we would've won."

Gunmen killed at least 34 Orthodox Christians in Endato village in Ethiopia’s East Arsi Zone on Thursday, November 5. It is the third mass killing of Orthodox Christians in the zone in 16 days, and the combined toll from the three attacks now stands at no fewer than 94. No group has claimed Thursday’s attack. The attack came three days after the regional government announced 73 arrests in connection with the October killings. What Happened Around 11 AM local time on Thursday, November 5, armed attackers hit Endato Kebele in Robe Woreda, in the East Arsi Zone of Ethiopia’s Oromia Region. A kebele is the smallest administrative unit in Ethiopia, roughly the size of a village, and a woreda is a district. The attackers targeted Orthodox Christian residents. Several people were wounded, and shops and homes were burned or damaged. The Ethiopian outlet Meseret Media reported that security personnel who tried to restore order were also caught in the violence. The bodies of 34 victims were gathered at Endato Debre Mitmaq St. Mary Church. The community warns the toll may rise as recovery continues. The names of the dead have not been released in the reporting available to us. Eyewitnesses say the attackers blocked roads into the area and cut off access to medical care. That claim has not been independently confirmed. We have coffee and tea now! This is what we are drinking in the CWT household. We hope you like it. The October Attacks Thursday’s killings follow two mass attacks in October. On October 20, attackers killed 14 people in Gora Silingo, Angada Kebele, in Merti Woreda. Four days later, on October 24, attackers killed 46 people in the villages of Tereta, Chole and Birbo in Shirka Woreda. Those figures come from the Holy Synod of the Ethiopian Orthodox Tewahedo Church, which reported more than 20 people seriously wounded and about 20 homes burned. At least five of the dead were children. Residents of Shirka told the Ethiopian newspaper Addis Standard that the October 24 attackers went house to house and used heavy weapons. Residents also said local authorities had been warned of a possible attack before October 24 and did not stop it. That claim has not been independently verified. Adding Thursday’s 34 dead, open-source reporting puts the combined toll at no fewer than 94 Orthodox Christians killed in less than three weeks. Who Is Responsible No group has claimed the November 5 attack, and no one has been publicly named. For the October attacks, the Oromia regional government blamed the Oromo Liberation Army (OLA), an armed group that has fought the Ethiopian federal government for years. The government refers to the group as “Shane.” On Wednesday, November 4, the Ethiopian Human Rights Commission (EHRC), a publicly funded body that operates independently of the government, also attributed the October attacks to the OLA. The OLA has denied attacking civilians in past incidents and had not publicly responded on the October attacks at the time of the most recent reporting. The EHRC finding covers the October killings only. It does not by itself establish who carried out Thursday’s attack in Endato. Arsi Zone is majority Muslim, and Orthodox Christians there are a minority living among Muslim and Protestant communities. The people being killed in these attacks are almost entirely Orthodox Christians, and many are ethnic Amhara. Open Doors reports that the attacks in the zone have primarily targeted Orthodox Christians and Orthodox churches, and that only the Orthodox community there reported incidents of this kind during the same period. In the February 26 attack, local sources said the gunmen entered a church shouting religious chants before opening fire on the people inside. The OLA is an ethnic Oromo nationalist group that recruits from both Muslim and Christian Oromo, and no Islamist group has claimed or been blamed for the Arsi attacks. The Ethiopian Orthodox Tewahedo Church describes the attacks as religiously motivated and aimed at its members. The Ethiopian Inter-Religious Council, a body of Christian and Muslim leaders, has said the attacks are not driven by religion or ethnicity. Arrests and Government Response The Oromia regional government announced that 73 people had been detained in connection with the October 24 attack in Shirka. The government has not said whether any of them have been charged or what evidence links them to the killings. Residents told Addis Standard that some of the people detained had buried victims, attended funerals, or helped carry and treat the wounded. Addis Standard could not verify the full number of detainees or the circumstances of each arrest. The Longer Pattern East Arsi has seen repeated killings of Orthodox Christians for years. In July 2026, the Ethiopian Human Rights Council (EHRCO), a private rights group, reported that at least 432 civilians and four local militia members had been killed in targeted attacks across East Arsi since 2021, with Orthodox Christian communities hit hardest. EHRCO counted 47 of those deaths in Robe district, where Endato sits, along with 156 people abducted there. More than 14,000 people have been forced from their homes across the zone. The attacks this year did not start in October. On February 26, attackers stormed a market in East Arsi and killed 20 Orthodox Christians and a Muslim guard, and local sources reported a priest was among those killed inside a church. Between May 30 and June 1, attacks in Aseko Woreda killed civilians, drove hundreds of residents into the forest, and burned the 101-year-old Telata St. Gabriel Church to the ground. What It Means for Church Safety Teams An armed group raiding villages is a long way from the threat most American churches face, and the pattern in East Arsi still applies to any church safety team. The attacks came in waves, with the third landing while families were still burying the dead from the first two. In Shirka, the people who showed up to bury victims and carry the wounded ended up exposed again. In Endato, eyewitnesses say blocked roads kept medical help from reaching the wounded. A church that has had one incident, or is holding a funeral or memorial after one, is at higher risk of a second, and the safety team is on its own until police and medics arrive. Ethiopian Orthodox parishes in the United States may hold memorial services and vigils for the victims in the coming weeks. Those gatherings will draw larger crowds than a normal service, and their safety teams should staff them at the level of their largest Sunday. A direct physical threat to American churches from the fighting in East Arsi is unlikely. What Your Safety Team Can Do Walk the outside of the building before every service. Check the parking lot, every exterior door, landscaping, dumpster and trash areas, and note anything that was not there last week. Get a named contact at your local police department. Meet that officer or supervisor in person, put the direct phone number in every team leader’s phone, and give the department a copy of your floor plan. Measure your police and fire response times. Ask your police and fire departments for average response times to your address on a Sunday morning and on a weeknight, write them in your team’s emergency plan, and plan to hold the building until they arrive. Keep full coverage up after any incident or credible threat. Do not pull people off posts to handle the first scene. Staff funerals, memorial services and vigils that follow an incident at the same level as your largest Sunday service. Plan for blocked exits. Show ushers and greeters a primary and a secondary exit for every room, and walk both routes with them before the next service. Stage bleeding control kits and train the people who will use them. Put tourniquets and hemostatic gauze where your team can reach them in under a minute, and get your team through a bleeding control class. Confirm casualty numbers and claims before you share them. Pass along figures from an identified source, and correct them when the source updates. What the Bible Says “Remember those who are in prison, as though in prison with them, and those who are mistreated, since you also are in the body.” (Hebrews 13:3, English Standard Version (ESV)) The writer of Hebrews was speaking to Christians who had already been through persecution themselves. Earlier in the letter he reminds them that they had been publicly exposed to reproach and had their property seized, and that they stood beside others who were treated the same way. In chapter 13 he tells them to keep remembering the ones still suffering, and he gives the reason: “since you also are in the body.” The believers in Endato, Merti and Shirka are part of the same body we are. They worship in a different tradition and a language most of us do not speak, and that does not change the command. Remembering them is a duty Scripture gives the church, and it starts with knowing what happened to them and praying for them by name when their names are released. “If one member suffers, all suffer together; if one member is honored, all rejoice together.” (1 Corinthians 12:26, ESV) Paul wrote this to a church that was divided and competitive, with members who thought some parts of the body mattered more than others. His answer was that the body cannot work that way. When one part is hurt, the whole body feels it. A village in Ethiopia that buries its dead at St. Mary Church for the third time in three weeks is part of our body, and we should feel it. For a church safety team, that suffering also carries information. The families in East Arsi have learned at great cost how attacks come in waves and how help gets cut off, and our teams should learn the same lesson and prepare our own congregations. Jesus warned His disciples in Matthew 24:42-44 that a homeowner who knew when the thief was coming would stay awake and keep his house from being broken into, and He told them to stay ready because they did not know the hour. The people of Endato had no warning on Thursday morning, and their loss is a warning to every church safety team that reads this article. For our teams, staying ready means walking the building before every service, knowing our response times by measurement, and training our people. Pray for the families in Endato, for the wounded, for the Orthodox Christians of East Arsi, and for those burying their dead again. Details may change as reporting develops.