In this paper we explore the hypothesis of -In this paper we explore the hypothesis of an "ad hoc horizon" for wireless multihop networks. This horizon is defined in terms of number of nodes and number of hops beyond which TCP based network services are not useful anymore. We show that with IEEE 802.11 this horizon is as narrow as 2 to 3 hops and approximately 15 nodes. Our results are obtained through simulations by choosing optimal settings (e.g., enabeling link layer feedback but no mobility) and focusing on worst case figures. The main reason for this horizon to form is TCP capture. We suggest that only combined efforts on MAC, routing and TCP level will permit to extend this limit, which can serve as a benchmark for future proposals. The ad hoc horizon also defines a narrow target size for the design of ad hoc routing protocols.