Neil Flanagan recently wrote about current and past proposals to heal the urban-renewal scars that separate the Kennedy Center — which should be a terrific urban amenity — from the city around it.
The KenCen, along with the Watergate complex and what’s now the Saudi embassy, stands in a tiny island isolated from both the city and the river by two parallel highways. Neil’s post focuses on a long history of proposals to bridge the chasm of I-66, built alongside this island as part of the grand urban renewal scheme that obliterated Foggy Bottom’s industrial heritage.
Yet the 1920s-era Rock Creek Parkway that runs on the riverfront through this stretch is perhaps a greater urban offense. It’s a limited-access highway that squeezes strolling pedestrians and cyclists into a narrow riverfront strip. It intervenes between the bike path and the river at one point, creating a particularly confusing, and dangerous, joint in the otherwise admirably complete trail network along the region’s waterways, and pretty much completely interrupting any pedestrian flow between the Mall and the waterfront. (Speaking of harrowing junctions, its at-grade intersection with I-66 creates a terrifying two-stage left turn at the end of I-66’s Independence Avenue ramp.)
And it could be eliminated with just two ramps — the ones shown in red in the map, linking the existing and underused ramps that link I-66 Extension to the Whitehurst Freeway, to Rock Creek Parkway. Adding these two ramps would enable cars that currently use Rock Creek Parkway to use the woefully empty I-66 that runs just two blocks east — and thus permit depaving Rock Creek Parkway (in pink), south of Virginia Avenue and north of the Lincoln Memorial.*
(A 1998 FHWA study also proposed the same ramp at the northeast quadrant of the interchange, but instead of a loop suggested a signal and a left exit. It also proposed to leave RCP, and grade-separated the Ohio/RCP intersection.)
Creating a linear park along the river between the Thompson Boat Center and the Lincoln Memorial would more clearly link three great linear open spaces — the Mall, Rock Creek Park, and Georgetown Waterfront Park and the upstream parks. (A clearer, perhaps grade-separated walkway behind the Lincoln Memorial would still be needed.) It would finally connect the KenCen and Watergate to the water, and break apart the asphalt chains that encircle the old Watergate Steps. It would also attach this little urban island to the city (well, Georgetown).
It would accomplish these aims at a cost far lower than decking over I-66, a proposal that has failed several times for want of funding. The surrounding renewal-era fabric would require retrofitting if such a deck were built, since most of it was built with high walls that ignore I-66.
Yes, direct access between Memorial Bridge and Rock Creek Parkway would be eliminated. Drivers would instead have access to the Roosevelt Bridge, which is currently denied, and could use Virginia and 23rd to reach Lincoln Circle and thus Memorial Bridge.
A similar concept was floated on POP in 2012 and roundly criticized in the comments, although it seems most of the commenters misunderstood which segment of RCP was being referred to.
* On second glance, the north Lincoln Memorial loop may be needed to allow Independence Avenue traffic to flow onto Memorial Bridge.
west north 
I’ve been working on this since Neil posted his article. Yours is basically a bare-bones concept, which is likely doable, but there’s a few things to note: even with more direct ramps between RCP and I-66, I don’t think you could depave RCP entirely. You’d still need to retain some lower access to the Kennedy Center. That said, you could likely do a 4-to-2 along that stretch of the RCP, and utilize the remaining space for on-street parking (which would further calm traffic) or an improved bikeway.
As you noted at the end, the 2nd Memorial loop would need to be retained for Independence-to-bridge travel. The connection between RCP and Ohio Drive would IMO also need to be retained, but could be reconfigured for a single southeast-bound lane.
The loop ramp you suggest would be troublesome. Because of the narrow right-of-way available, it’d be a pretty tight loop. In order to draw the most traffic off of RCP, it would need to be 2 lanes and also would require widening the existing ramp from the Whitehurst to 66. The NPS parking lot it would take out is why I’d keep RCP as 2 lanes with on-street parking (though you probably see differently on this than I do).
The northbound ramp would also be troublesome. It’d be a very tight fit to get it in and avoid impacting the old Lime Kilns, though it might be possible.
I suggest sending 66 to RCP through the open space between Virginia Avenue and K Street and west of 27th Street. End it with a signalled intersection at RCP. Inbound Whitehurst Freeway traffic could continue to enter RCP through 27th Street. Virginia Avenue traffic would enter RCP as it does currently and meet 66 at the new intersection. Whitehurst Freeway no longer would connect to Virginia Avenue through 27th Street.
Inbound Whitehurst Freeway traffic would get to Virginia Avenue through the new intersection at 66 and RCP. Outbound K Street would continue to use the exit ramp on the right to RCP.
Yes, my intent was to propose a simpler solution that improved the area at a cost that wouldn’t reach nine or ten figures.
The nice thing about implementing changes to RCP, rather than to I-66, is that any new ramps could be very tight rather than Interstate-spec. I do think that removing the Thompson parking lot is a feature, not a bug, since I’ve had a few conflicts with drivers crossing the trail to exit that lot.
I regularly ride along Ohio Drive in West Potomac Park, and its “park drive” configuration always seems confusing. If any more two-lane sections are built, a chicane-like arrangement with pockets of angle parking might keep the speeds down and mitigate the “wall of parked cars along the river” effect that Ohio has.