There is no longer one FAR.
The biggest rewrite of federal acquisition rules in forty years did not arrive as a rule. It arrived as fifty-three deviations, adopted agency by agency, on staggered timelines, before a single final rule existed. The text you are being evaluated against is in the solicitation, not in the regulation you can look up.
What actually happened
Executive Order 14275, signed April 15, 2025, directed that the FAR contain only what statute requires or sound procurement demands. What makes the overhaul unusual is not the ambition. It is the sequencing.
Normally a change of this size moves through notice and comment, gets finalized, and then takes effect. This one ran backward. The FAR Council rewrote all fifty-three parts as model deviation texts, posted them to acquisition.gov rather than the Federal Register, and gave every executive agency thirty days to adopt each one through its own class deviation. The new rules went operative first. The rulemaking came second, and is still running.
Phase two began in June 2026 with the first four proposed rules, covering twenty parts. Parts 6, 7, 10, 18, 26, 37, and 41 published June 23 with comments closing July 23. The rest are still in the pipeline: the bundles covering Parts 34, 42, and 48, Parts 9, 27, and 47, and Parts 16, 17, and 35 remained open cases into August, cycling between the Office of Federal Procurement Policy and the Office of Information and Regulatory Affairs for review. On the defense side, DFARS class deviations were still being revised as recently as August 5, including a second revision to Part 6.
The system of record is wrong on purpose
This is the part that costs people money, and almost nobody has been told about it directly.
GSA's deviation instructed its workforce to follow the overhaul text rather than the codified FAR, and then said the quiet part in writing: system updates may lag policy updates. Contracting officers were told to rely on the deviated provisions in the solicitation even when SAM.gov prompts for representations that no longer exist. The Department of Commerce repeated substantially the same warning in April 2026, which tells you the lag is not a startup problem that resolved itself.
The regulation moved. The system of record did not.
If you build proposals off auto-generated clause lists, SAM.gov prompts, or a template that worked last cycle, you are now working from a source the government itself has told its own people to ignore. The operative text is whatever the solicitation says it is, and it varies by agency. Two agencies buying the same service in the same month can be running different versions of the same FAR part.
Your certification lost its place in line
The Rule of Two survived, which is what most of the trade coverage led with. The more consequential change got less attention: there is no longer an order of precedence among the small business socioeconomic programs. Contracting officers are no longer required to prioritize HUBZone, SDVOSB, WOSB, or 8(a) set-asides over a general small business set-aside.
For task orders under multiple-award contracts, the set-aside requirement also moved from mandatory to discretionary. Read those two together and the picture is clear enough: your certification still qualifies you, but it no longer moves you up the queue automatically. Discretion replaced sequence, and discretion favors the firm the contracting officer already knows.
Two things cut the other way. The 8(a) incumbency limitation on SDVOSB sole sourcing was removed, so an agency can now make a sole source SDVOSB award even where an 8(a) firm currently performs the work. And 8(a) follow-on work can now be set aside for HUBZone, SDVOSB, or WOSB without a formal SBA release, which quietly unlocks a category of recompetes that used to be closed. Size and status protest procedures are otherwise unchanged.
Worth holding alongside all of this: the statutory SDVOSB prime contracting goal went up, from three percent to five, in the FY2024 NDAA. The demand signal strengthened while the regulatory priority weakened. Those are not the same lever, and the gap between them is where the work is.
The rule that changed under the rule
The model deviation for Part 1 introduced a regulatory sunset: any FAR provision not required by statute would expire after four years unless the FAR Council renewed it. That was widely reported in late 2025, and it is the version most summaries still carry.
The proposed rule softened it. Automatic expiration is gone. Sections do not lapse on their own and can only come out through rulemaking; instead the Council intends to run periodic sunset reviews inviting public comment on what should be allowed to expire.
Take the lesson rather than the detail. Anything you read about this overhaul has a shelf life measured in months, including this note. One live example: the Rule of Two above the simplified acquisition threshold is not statutorily mandated, which makes it exactly the kind of provision a future sunset review could reach.
What a small shop should actually do
Not "retrain the team." You are the team. Three things, in order of payoff.
Add a deviation check to every opportunity assessment. Before you write anything, find out which parts that specific agency has adopted and when. It is a five-minute question that changes which clauses you are agreeing to.
Read the solicitation as the authority. When the solicitation and the codified FAR disagree, the solicitation wins right now. When your registration system asks for something the solicitation dropped, the solicitation still wins.
Stop competing on certification alone. With precedence gone and discretion expanded, past performance, a contracting officer who recognizes your name, and a capability statement that survives a thirty-second skim matter more than they did eighteen months ago. There is one bright spot here: the Part 4 proposed rule cuts roughly half the information required to register in SAM.gov, which lowers the barrier for new entrants. That helps you and it helps everyone competing against you.
Why this is a data problem now
The old FAR was hard but stable. You could learn it once. The new one is simpler to read and harder to track, because the authoritative version is distributed across agency deviations that move independently and update without announcement.
That is a monitoring problem, not a legal one, and it is the kind of thing a small shop loses by attrition rather than by any single mistake. Nobody misses a deviation dramatically. They just quote against last year's clause set, agree to something they did not read, and find out at award or at audit. The firms that come through this well will be the ones that made tracking a routine instead of a scramble.
Full disclosure: we are an SDVOSB. Everything in section 03 lands on us the same way it lands on you, and we would have preferred the precedence stay. That is a reason to read this carefully, not a reason to discount it. Where the overhaul is good for small business, we said so.
Rules that change quarterly across a dozen agencies are a tracking problem, and tracking problems are what automation is actually good at. A Recon maps what your team is watching by hand and what it is missing.
Start with a Recon