Good morning. It is Thursday, June 4th, 2026, and this is your Morning Brief.

COHOST: The useful theme this morning is that several stories that look like relief at first glance are not actually relief once you look at who still has to absorb the cost or make the workflow work.

HOST: Exactly. Stronger services activity is not the same thing as easier rate cuts. More homes for sale around Columbus are not the same thing as lower prices. New mortgage tools do not help if a lender's front line does not know how to use them. And new AI models are not just about who is smartest anymore. They are about where the model runs, what it costs, and how much governance overhead it creates.

HOST: Start with the national signal that matters most before Friday's payrolls report on June 5th. The Institute for Supply Management said its May services index rose to 54.5 from 53.6 in April. New orders improved to 57.3 and business activity rose to 57.7, both signs that a huge part of the U.S. economy is still expanding at a decent clip.

COHOST: But the part that keeps this from sounding clean is under the surface. Hiring is still soft and prices are moving the wrong way.

HOST: Right. ISM's employment index fell to 47.9, which means service firms on balance are still cutting or freezing hiring. At the same time, the prices index rose to 71.3, its highest reading since August 2022. Survey respondents pointed to fuel, transportation, construction materials, software, and supply constraints tied partly to the conflict in the Persian Gulf and to tariff dynamics.

HOST: Put that together with Wednesday's ADP report, which showed private payroll growth of 122,000 in May, and the message is not recession but friction. Demand is holding up better than many businesses would like if they are waiting for lower rates, while labor demand is strong enough to delay relief but not strong enough to feel easy for workers looking for momentum.

COHOST: The practical consequence is largest for borrowers, rate-sensitive businesses, homebuilders, and employers trying to decide whether this is a pause or a re-acceleration.

HOST: Exactly. If Friday's payrolls report also comes in firmer than expected, Treasury yields could stay elevated and keep mortgage relief out of reach for longer. If payrolls disappoint, then the cleaner read becomes that businesses are still spending and ordering, but are hesitating on headcount and getting squeezed on costs. The counter-signal here is that services strength is real. Seventeen industries reported growth in May, and only one reported contraction. So this is not a broad collapse story. But it is also not a low-inflation, easy-cut story. Watch Friday's payroll count, unemployment rate, and wage growth. That is the next piece that decides whether this morning's data means resilience or just sticky pressure.

HOST: The second major story is local, and it cuts against what a lot of Central Ohio buyers probably hoped would happen this spring. Axios Columbus reported Wednesday that home inventory is rising and homes are sitting longer, yet prices are still moving up across most of the region. Columbus Realtors' April housing report showed closed sales down almost 14 percent from a year earlier, while new listings rose 7.8 percent and inventory rose 13.4 percent.

COHOST: Normally that sounds like the setup for relief. More supply, slower sales, longer listing times. But the price break still is not here.

HOST: Exactly. In Franklin County, the average sale price still climbed 5.9 percent from a year ago to just over 391,000 dollars. That is the important new signal. The market is becoming less frantic, but it is not becoming cheap. For listeners in Columbus and the surrounding counties, that means more choice does not yet mean better affordability. Buyers may get a little more room to negotiate inspection timing or seller concessions, but monthly payment math is still being controlled by price levels and mortgage rates, not by a sudden collapse in demand.

HOST: There is also a class signal inside the local data. Axios noted that more of the apparent strength is coming from affluent buyers and suburban demand, while lower-cost relief remains limited. That means the market can look healthier in averages than it feels for first-time buyers or moderate-income households trying to buy inside Franklin County without prior-home equity.

COHOST: So the counter-signal is that this is still a normalization story, not a fresh bubble story.

HOST: Right. Slower sales and more inventory do matter. They tell you the market is less one-directional than it was in the most overheated years. But the fresh point is that Central Ohio still appears undersupplied enough, and high-income demand still resilient enough, that prices are not giving back much ground. The next watch item is whether May and June local data finally show price compression in Franklin County and the inner suburbs, or whether more inventory just gets absorbed by buyers who can still pay up.

HOST: On home lending and mortgage, today's useful angle is not simply that rates are high. It is that lenders are operating in a market where every technical edge matters because the macro backdrop still refuses to get easier. Freddie Mac's weekly survey, published on May 28, still shows the average 30-year fixed mortgage at 6.53 percent, up from 6.51 percent the week before, with the next survey due today, Thursday, June 4th, at noon Eastern.

COHOST: Which means this morning's mortgage story is partly about waiting for rates and partly about what lenders can control before rates improve.

HOST: Exactly. The controllable part is getting borrower eligibility and pricing right. Fannie Mae said its 2026 area median income limits will take effect on June 13 in Desktop Underwriter, Loan Delivery, and its AMI lookup tools. That matters because AMI thresholds determine eligibility for HomeReady, RefiNow, Duty to Serve programs, and certain loan-level price-adjustment waivers. In plain language, the file date and the AMI lookup are becoming more commercially important.

HOST: Then layer on FHFA's April move into what it called a new era of credit-score competition. Fannie and Freddie are already accepting VantageScore 4.0 loans from approved lenders, while the Enterprises still expect historical FICO 10T scores to be published this summer. That does not mean every lender suddenly gets a huge new borrower pool overnight. It does mean the firms that understand score-model choice, AMI-based pricing relief, and borrower communication can defend conversion better than firms still selling only a headline rate.

COHOST: The operational consequence is clear. In a forgiving market, weak execution is annoying. In a six-and-a-half percent market, weak execution is lost revenue.

HOST: That is the point. Borrowers need a lender that can explain whether they qualify for a waiver, whether a score model could help, and whether waiting a week changes the economics of the file. The counter-signal is that none of these tools solves the big affordability problem by itself. If payrolls are strong on Friday, bond yields can still undo a lot of that progress. But between now and June 13, lenders have a very concrete to-do list: train the front line, tighten pricing conversations, and prepare for another week where macro relief may not arrive on schedule.

HOST: On AI and technology, the fresh model story is less about a single winner and more about the market splitting into two useful directions. Google announced Gemma 4 12B on Wednesday, June 3rd, calling it a unified multimodal model designed to bring agentic AI directly to laptops. Google says it can run locally with 16 gigabytes of memory, includes native audio input, and is released under Apache 2.0.

COHOST: That is a very different message from the biggest frontier models. It is not just "more intelligence." It is "bring capable multimodal AI much closer to the workstation."

HOST: Exactly. For enterprises, especially in regulated environments, that is a meaningful shift. A model that can run locally or near-locally changes the privacy and deployment conversation for document review, call analysis, research assistants, and internal copilots. It does not eliminate governance, but it can reduce some exposure that comes with sending every workflow to a remote frontier model.

HOST: At the high end of the market, Anthropic's Claude Opus 4.8 remains the more obvious frontier benchmark for coding and long-running agentic work. Anthropic says the new version keeps the same price as Opus 4.7, adds stronger judgment for agentic tasks, and offers a fast mode that works at two and a half times the speed while being three times cheaper than before. That matters because the enterprise AI question is increasingly about routing. Local or lighter-weight models for privacy-sensitive and repeatable tasks. Frontier models for harder reasoning, coding, orchestration, and review.

COHOST: And OpenAI's latest visible signal this week is simplification, not another splashy model launch.

HOST: Right. OpenAI's June 2 ChatGPT release notes said ChatGPT will retire OpenAI o3 on August 26, 2026 and GPT-4.5 on June 27, 2026 in the product, while keeping API access unchanged. The important read is that major vendors are trying to reduce lineup clutter at the same time customers are trying to reduce model sprawl. Procurement, risk, and engineering teams do not want fifteen overlapping choices. They want a smaller set of models they can test, route, monitor, and govern.

HOST: The counter-signal is that smaller local models still trade off raw capabi...