I think you refer to pruned node, which download whole blockchain but only store some latest blocks and UTXO set (which necessary for verification purpose). People already can run pruned node since some years ago using Bitcoin Core software.
You know ABC, basically just think we need this brainstorming to meet more and more "Core Wallets" out there among BTC hodlers, instead of hardware and SPV wallets. [1]
We don't live in ideal world though. Not everyone bother put resource or handle complexity running wallet backed by their own full node.
so prune nodes are very close to the concept if they save ALL unspent transactions (non-oxidized transactions) - which amazingly this could also provide potential compatibility with GDPR principles in future [2].
Prune node actually save all UTXO[1], not all TX. Otherwise, size of chainstate folder (created by Bitcoin Core) would be much bigger.
I also do follow ideas like assumeUTXO which shows us there are still some sort of improvements needed to get considered[3]:
"Rather than the status quo — setting a number of blocks and compressing historical blocks prior to that milestone — O’Beirne’s assumeUTXO is an experimental way for new Bitcoin full nodes to delay their need to verify historical transactions until the user receives recent transactions."
now look at this oxidation fee how forces the entire network and hodlers to achieve better performance of prune-nodes in bitcoin blockchian and facilitate ideas like assumeUTXO.
Assume UTXO also already exist on Bitcoin Core[2], but works differently than many of us expect. I also don't understand how oxidation fee improve prune node, since based on my understanding it doesn't reduce total UTXO.
[1]
https://learnmeabitcoin.com/technical/transaction/utxo/[2]
https://github.com/bitcoin/bitcoin/blob/master/doc/design/assumeutxo.md