Qubic

A weekend, human look at the team relentlessly building behind the scenes of Qubic, prompted by how candid the last Tech on Deck session was. Building core Qubic is not typical software work. Code that goes live on the network and fails does not just log an error. It has real consequences. A few things the core developers were honest about: 🔹 The hard start. One described their first release stalling after a single tick, a tiny bug that took a long time to find, and the doubt that came with it. 🔹 The environment. They write in C++ without the standard library most engineers rely on, so foundational pieces are built and tested in-house. Smart contracts run inside a strict sandbox with invariants the code must honor. 🔹 AI’s role. AI is a helpful second pair of eyes for small bugs, but limited because it does not know the codebase’s hidden rules. It will flag non-issues while missing real ones. 🔹 The process. Community contributions cannot merge directly. At least one core dev reviews each change, and two review more complex code, under a “four-eyes” rule the team applies to its own work too. One developer’s highlight: Building the contract-verification tool and seeing it run on the first real community smart-contract submission. What comes through is a team that treats a live network with the seriousness it deserves. Methodical, careful, and aware that real value depends on getting it right. → https://github.com/qubic

The June 11 Qubic All-Hands packed a lot into one session. The short version: the board approved a fiat on-ramp and a crypto payment card covering 173+ countries. The scientific team got Neuraxon V2.0 paper accepted at AGI-26 in San Francisco. Core tech locked in four protocol milestones between now and August, including outsourced computing on mainnet and the next emission halving. There's more in there on BD pipeline numbers, a new community lead, a marketing leadership transition, and regional expansion into LATAM and APAC. Full recap with all the context and links:

Most blockchain fee models redistribute value. Qubic's destroys it. On most networks, your transaction fee goes to a validator, a sequencer, or a treasury. On Qubic, every fee is burned, permanently removed from supply. The recent Tech on Deck session confirmed a third mechanism joining that design: • Smart contract execution: each contract run pays a fee that is burned. • Oracle calls: each request for outside data is burned. • Outsourced Computation: each off-chain action a contract triggers will pay a fee that is burned. If you know Ethereum's EIP-1559, the concept is familiar. The difference is scope: Not a partial burn on one fee type, but every fee, from every mechanism, every time. On Qubic, network activity is subtraction. Usage does not just generate throughput metrics. It permanently reduces what exists. Today, weekly emissions still exceed what usage burns. But emissions were just cut in half by an on-chain vote of the network's 676 Computors, effective epoch 227, and every new contract deployment, Oracle query, and outsourced instruction pushes the other side of the equation. At some point those curves cross, and the network turns net deflationary. That milestone is not on a roadmap. It is a function of usage, which means it gets earned, not announced. How the burn architecture works → https://lnkd.in/ewYeVF_e

Join us Thursday, June 11th at 3:00 PM UTC for the Community All-Hands AMA This is your front-row seat to everything happening inside Qubic right now. RSVP to save your seat and get a reminder the moment we go live. See you Thursday.

Once a smart contract can send an authorized instruction off-chain, a category of applications opens up that was previously out of reach on Qubic. From the recent Tech on Deck session, the two clearest examples: -> Cross-chain action. A Qubic contract can trigger a transaction on Bitcoin, Ethereum, or another chain, wherever Qubic runs a dedicated Outsourced Computation Machine pointed at it. The decision happens on Qubic, the machine carries the instruction out, and the action executes on the other chain, with no manual intermediary. -> Custody handovers. Control of a shared multisig wallet on another chain can pass from one signer set to the next, triggered by a contract and approved by the network. The trust model underneath is what makes it work. An instruction only leaves Qubic after 451 of the 676 Computors independently sign it, and the receiving system can verify those signatures before acting. The outside world is not asked to trust Qubic. It can check the network’s approval directly. This is in active development, with go-live targeted for July 29 and public testing before then. → https://lnkd.in/gy9_dEKV

We've been measuring artificial intelligence wrong. And we might not even have the right tools to fix it yet. The Qubic Scientific Team just published The Neuraxon Intelligence Academy, Volume 10, and this one tackles a question the AI industry prefers to speed past: How do you actually measure the intelligence of a machine? Here's the uncomfortable part. The g factor, the closest thing psychology has to a universal measure of cognitive ability, was never designed to describe an individual. It describes variance across a population. When researchers apply the same factor analysis to large language models, they find a clean, dominant first factor. Looks like proof of general intelligence. But models trained on similar data, with similar architectures, will naturally correlate across tasks. What we may be extracting is a "factor of scale" dressed up as cognition. Then there's the contamination problem. Benchmarks like MMLU and GSM8K are published online. LLMs are trained on the internet. Fragments of the test end up in the training data. Empirical audits have found contamination levels as high as 45%. When researchers rewrote leaked questions without changing the difficulty, accuracy dropped by over 20%. François Chollet's ARC-AGI benchmark was built specifically to resist this. ARC-AGI-3, launched earlier this year, drops AI agents into interactive environments with no instructions, no natural language prompts, and no way to memorize the answer. Humans solve 100% of them. Frontier AI models score below 1%. The full article traces the line from Spearman's 1904 positive manifold through IQ psychometrics and into the current state of AI evaluation. It also connects those ideas to the direction Neuraxon and Qubic are taking with brain-inspired, temporally dynamic architectures. Worth the read if you care about what "intelligence" actually means when we say it about machines.

A common misconception about Outsourced Computation is worth clearing up. It does not mean Qubic is renting out compute power. It does not mean miners will run external workloads such as LLMs. As the core team put it directly in the last session: “Qubic does not outsource compute, it outsources a computation.” The distinction is the whole point: Renting compute is general-purpose. You get processing power and use it however you like. Outsourced Computation is narrow and specific. A smart contract authorizes one defined action, such as moving funds or triggering a transaction on another chain, and sends that single instruction outward for an external system to execute. The control that makes this safe: Nothing leaves the chain until 451 of the 676 Computors independently sign the request. Those signatures also let the receiving system verify the network genuinely authorized the action, rather than trusting a single party. In short, this is not a compute marketplace. It is a smart contract acting on the outside world with one instruction the network has collectively approved. → https://lnkd.in/gy9_dEKV

For most of its life, a Qubic smart contract could reason about anything but act on nothing outside the chain. Outsourced Computation closes that gap. A useful framing, drawn from the recent Tech on Deck session: Earlier this year, Oracle Machines gave smart contracts eyes and ears. A contract can request a verified fact from the outside world, a price or an event outcome, and receive it on-chain. Information flows in. Outsourced Computation reverses the flow. A contract sends an authorized instruction out to an external system, which executes it. Intent flows out. The two are mirrors. One brings the world in. The other reaches into it. Outsourced Computation is the third and final pillar. Combined with the smart contract itself, the system is complete: A brain that decides. Senses that observe. And hands that act. A contract that can do all three is no longer just a contract. It is the foundation of an on-chain agent, one that operates beyond its own chain. The work is in active development, designed with Qubic founder Come-from-Beyond, with go-live targeted for July 29. Not live yet, but the path is dated and scoped. How it works under the hood → https://lnkd.in/gy9_dEKV

A relaxed weekend recommendation for the technically curious.   The Qubic Mining Academy has been updated for the Dogecoin integration. It is a free, step-by-step, easy to understand course, and it now covers how to point an ASIC miner at a Qubic pool, alongside the fundamentals.   What it walks through:   * What your mining hardware is actually doing * How a Qubic pool works and where rewards come from * Practical settings and best practices so you are not wasting power * No prior mining experience required The distinction worth keeping in mind: mining on Qubic is not throwaway computation. The same work that earns rewards contributes to the network's AI training, so the energy produces something rather than simply proving effort was spent.   Dive in and start the course: https://lnkd.in/eS8TGrDV

SpaceX’s IPO filing, made public last month, is worth a read as a market signal. The S-1 gives investors a formal way to price exposure to three areas the crypto market has been circling for years: ● A Bitcoin position of 18,712 BTC disclosed on the balance sheet ● A payments and banking push inside X ● A significant move into AI compute and data-center capacity What stands out structurally is that these remain three distinct business lines under one corporate roof, each managed separately. Qubic was designed the other way around. The mining that secures the network is the same computation that trains AI. There is no separation between a crypto operation and an AI operation, because Useful Proof of Work is a single mechanism that produces both at once. The convergence that major companies are now assembling piece by piece is the premise Qubic started from. Learn how Qubic is leading the upow race: https://lnkd.in/euk8FRC4