Tomislav Grospić

The 2026 Web3 inflection point won’t be faster blockchains, but trustless collaboration between AI agents. Today, the execution model of most blockchains is still based on: instant short transactions, ordering, and then sequential or parallel execution with synchronous calls. This model is almost unusable for an agent economy. A blockchain that can truly support an agent economy needs: ⚡️ Coroutine-style transactions (pause / wait / resume + ACID) ⚡️ Process orchestration primitives (join, select, async coordination) ⚡️ ACID cross-shard execution (true horizontal scalability) ⚡️ OCAP-native security (secure agent-to-agent interactions, preventing prompt injection attacks) When AI agents begin executing autonomously on-chain: Blockchains without a concurrency model will become the biggest architectural bottleneck, and systems without capability-based security will become the largest security vulnerability. A distributed concurrency control layer will be the key infrastructure for the Web3 × Agents era. $ALUX #Web3 #AI #Blockchain #AIagent

As part of a recent upgrade to our TupleSpace Virtual Machine (TVM), ALUX has introduced new behavioral channel types to enhance performance and predictability: ※ Single-Owner: A name (channel) is exclusively owned by one process at any time. ※ Affine: A name may be used at most once. ※ Linear: A name must be used exactly once. These constraints are enforced at runtime by our bytecode VM, enabling more efficient execution paths. Why it matters: ALUX is building a Pi/Rho calculus-based VM designed for composable, long-running concurrent transactions on-chain. To scale without compromising performance, we’ve introduced a class of channel types that unlock key optimizations. For example, Single-Owner channels support lock-free access and can be used to hold sequential VM sandboxes like the EVM. Affine and Linear channels ensure deterministic usage, reducing the overhead of recording extra information for validator replay. This is where formal theory meets practical engineering—ALUX brings it to life. $ALUX #Web3 #Blockchain #Concurrent #EVM

Belief = -Faith #21 Denotational Design - Conal Elliott youtube.com/clip/Ugkxlw1a_…

Dependent types == Reasoning #21 Denotational Design - Conal Elliott youtube.com/clip/Ugkx_dBvU…

High & low level languages. #21 Denotational Design - Conal Elliott youtube.com/clip/Ugkxto7UN…

🚀 在 ALUX 对前沿技术的探索中，一次具有里程碑意义的突破已经实现：EVM 与 Tuple Space 的初步对接取得成功。如同星舰计划的草蜢火箭成功跃迁，这一突破也初步验证了 ALUX 原创的在并发的进程演算虚拟机上支持 EVM 的可行性。在测试中，EVM 在沙箱环境中运行，遇到陷阱时会跳转至 ALUX 的 TVM （Tuple Space VM）核心态进行处理，处理完毕后再恢复至 EVM 用户态执行。这使得 EVM 能 “冻结” 在特定状态，等待其他操作完成后再继续执行。这第一次让 EVM 具备了链上并发执行需要等待的长时间交易的能力，极大地扩展了 Web3 技术的应用范围。 💡 接下来，ALUX 计划在 EVM 中实现对 RPC 或 REST 的调用，允许 EVM 在获取结果之前 “等待”，直到操作完成再继续执行。此举将大幅拓展区块链的应用场景。面向未来，ALUX 还将支持 WASM。 🌌 ALUX 引入的 Replay Log 应对了由于并发带来的不确定性，确保其他节点能够确定性地复现状态转移。从一开始，ALUX 就专注于优化这一机制，确保绝大多数的事件无需记录，以减少通信开销。 🛠 ALUX 基于 OCAP 的虚拟机模型能够同时支持 UTXO 和账户模型，通过这一创新，可以开展各种创意拼接，成为比特币二层网络的绝佳解决方案。 🖥 而更重要的焦点，ALUX 更是利用了 [并发] + [组合] 的特性连接 Web2，这是实现真实世界资产（RWA）在 ALUX Network 上运营的关键所在。无论在 Web2 和现实世界中，并发无处不在，因此 Web3 技术未来无疑将基于并发架构。正确的架构设计将引领 ALUX 探索无限可能，朝着星辰大海航行。 @alux_network $ALUX

Happy New Year! Here are the New Year's greetings from ALUX's founder, @atticbeeus

#rchain #rpc #rholang #blockchain #scalability

@IOHK_Charles @jack @elonmusk @jamesdouma @RadarMoron @JeffTutorials @karpathy Concurrency wins! Ask the master of computation @leithaus.

@jack @elonmusk @jamesdouma @RadarMoron @JeffTutorials @karpathy just wait until you guys discover haskell

Yo @elonmusk please add software release notes into the Tesla app. Would be nice to see what is new right from my phone.

@Sotoyour @leithaus @rchain_coop The most importantly, we are now in the very last tiny step of pushing block-merge code on test net.

@Sotoyour @leithaus @rchain_coop Hi @Sotoyour! I'm glad that you are interested in more details about block-merge necessary changes.

280 April 27 2022 Community Debrief – dappy presents “Easy RNode” blog.rchain.coop/2022/04/28/deb…

@Sotoyour @leithaus @rchain_coop #simple-2-partition-branching-and-merging" target="_blank" rel="nofollow noopener">github.com/tgrospic/rchai… Repo contains some nice DAG visualizations from the simulator with dynamic view of blocks with supported finalized fringe. As of my knowledge no other chain have this kind of view.

@Sotoyour @leithaus @rchain_coop You can see from my last PRs that I'm cleaning up existing implementation of single-parent and preparing to inject the new finalizer.

@Sotoyour @leithaus @rchain_coop Under the Block merge (Hard Fork 2) milestone we have 94 closed PRs. I'd love to schedule a meeting with whoever is interested and go down the rabbit hole for each of them and discuss any of them in details. 🦸 github.com/rchain/rchain/…

菩萨蛮·观禅 作者：@11Shixi 凭生玉雨消杂念，洞悉神曲沿风范。枯坐乍光年，明通在眼前。 平端吞浩瀚，空性任呈现。此处乃佛天，何须无限还。 #词牌 #古诗词 #菩萨蛮

While Deutsch argues that computation must ultimately be a physical theory it seems to me that every scientific theory is a computational theory.

@leithaus Thank you for the most inspirational thoughts on internet!

@leithaus How cen we prove that theory has infinite predictions? I'm trying to understand what is the difference between infinite theories and infinite predictions.

So, any scientific theory of reasonable utility is a theory that makes an infinite number of predictions which are accessible to computation, i.e. it offers a means to compute an infinite number of predictions. In other words, a scientific theory is a computable theory.

@leithaus One can argue that computation is just simbolic manipulation which does not have to be related to anything physical. Is "accessible to computation" has the same meaning as physical?

If it makes a finite number of predictions, it’s not a very interesting scientific theory. On the other hand, if it makes an infinite number of predictions, but those predictions are not accessible via computation, it is again not very useful.