Difficulty adjustment algorithms (DAAs) are a vital component of every Proof-of-Work (PoW) blockchain. They regulate mining difficulty which in turn modulates interblock times. Maintaining a stable block-production rate and consistent transaction throughput is crucial for the smooth operation of a blockchain and can positively affect its reputation. The challenging part of a DAA is that the actual blockchain hash rate is unknown. Therefore, difficulty adjustments must be based on historical data, estimations and/or predictions. For Bitcoin (BTC), things are relatively simple. The vast hash power of Bitcoin’s network makes it largely immune to fluctuations, allowing the difficulty to be adjusted infrequently (approx. every two weeks) without affecting throughput. On the other hand, the rest of the PoW blockchains, such as Bitcoin Cash (BCH), are more susceptible to hash power fluctuations and require a more adaptive DAA which adjusts the mining difficulty after every block. Such DAAs already exist, have been battle-tested and have proven effective at maintaining average interblock times close to the target. Nevertheless, to our knowledge, no DAA successfully addresses the problem of accumulated drift. In fact, both BTC and BCH halvings of April 2024 occurred almost 9 months ahead of schedule due to this drift. In this paper, we propose a novel yet very simple DAA based on a negative exponential filter that not only keeps the block-production rate stable in the long run but also eliminates any already-accumulated drift.