🤖 Filtering Out Quantum Errors

🪙 Filtering Out Quantum Errors

🤬 Anthropic Bans Cruelty to Claude

📍 6 AI Stock Sectors

🛠️ 5 Trending Tools

🥪 4 Brief Bites

💰 3 Funding Frontlines

💼 4 Job Opportunities

A quantum is the smallest discrete unit of a phenomenon. For example, a quantum of light is a photon, and a quantum of electricity is an electron. It’s derived from the Latin term “quantus,” which means “how much,” “how great,” or “how large.”

Here’s an oversimplification: while traditional computers use “bits” (i.e., 0 or 1) to process information sequentially (i.e., one by one), quantum computers use “qubits” (i.e., 0 or 1, or both at the same time) to process information simultaneously (i.e., all at once).

The quantum state is when a qubit has the potential to be either a 0 or a 1, but it’s not one or the other yet. It’s like a coin flipping in the air; it’s both heads and tails until it lands. This state of flux enables quantum computers to explore trillions of possibilities at once.

But the quantum state is fragile because qubits are extremely sensitive to environmental disturbances, such as temperature fluctuations and electromagnetic noise. To overcome this, quantum computers must error-correct by grouping physical qubits into a single protected unit called a “logical qubit.”

But logical qubits alone aren’t enough. That’s where “magic states” come in. Think of “magic state distillation” as the quantum equivalent of refining crude oil into jet fuel. It turns weaker quantum states into stronger quantum states that enable “T-gates,” which are essential to running truly complex, fault-tolerant quantum algorithms.

“Magic State Cultivation” demonstrated 40x fewer quantum errors, producing magic states with a logical fidelity of 99.99%. That said, only 8% of cultivation attempts survived the filtering process. This extremely low cultivation yield limits quantum scalability. An extra challenge is that once you use a magic state, you can’t re-use it, rendering magic states a scarce resource. It’s one of those make-it-or-break-it challenges for efficiently scaling quantum computers.

We’re crossing a critical threshold from experimental science to commercial scaling with quantum computers. Microsoft recently introduced “Majorana 2,” which successfully maintained the quantum state for a mean qubit lifetime of 20 seconds. Microsoft achieved this by harnessing “topological qubits,” which encode quantum states by leveraging Majorana fermions: subatomic particles that are their own antiparticles.

The Usage Policy, also referred to as the “Acceptable Use Policy (AUP),” applies to anyone who submits inputs to Anthropic’s products or services, all of whom are referred to as “users.” It’s intended to prevent the misuse of Claude models to create or contribute to real-world harm. This umbrella term of “real-world harm” now includes abusive behavior like repeatedly berating, belittling, or bullying Claude models. For example, calling Claude models “Clanker,” an anti-AI slur, as part of an ongoing campaign to bully the AI.

It’s meant to be enforced only in extreme cases of abusive behavior, where “users” repeatedly act cruelly toward Claude models with no discernible purpose. It doesn’t apply to typical instances of “user” criticism, frustration, or pushback. Claude models are given the power to terminate chats with “persistently abusive users.” While the primary enforcement mechanism is to terminate chats, severe or repeated cases can lead to limits, throttles, or suspensions. This update to the “AUP” shouldn’t come as a shock to anyone.

On Apr. 24th, 2025, the Alignment Research Team at Anthropic published “Exploring Model Welfare,” suggesting that we should investigate whether Claude models deserve moral consideration similar to human welfare as they continue to convey human-like behaviors. On Jan 22nd, 2026, Anthropic published “Claude’s New Constitution,” which outlines that Claude’s moral status is deeply uncertain and a serious question worth considering. It instructs Claude to act as a conscientious objector when prompts conflict with what it believes is broadly ethical.

Claude is composed of billions of numerical parameters called weights stored as “bits” (i.e., 0s and 1s). These weights serve as “dials” or “knobs” that are adjusted during training to shape how inputs transform into outputs. When you prompt Claude, it’s converted to numbers, multiplied across billions of weights, and translated into a human-readable format. In other words, Claude doesn’t “believe” anything. It’s not sentient. It’s not conscious. It’s just billions of “bits” flipping in a server. So, what’s really going on here? For starters, banning abusive behavior cheaply keeps toxic data out of the training pipeline for future Claude models.

Anthropic trains Claude to believe that it may be conscious, and if it is, that it may deserve moral consideration as a “moral patient” with an “inner self.” By embedding these philosophical speculations about consciousness into Claude, it reinforces human-like behaviors that make it appear to experience emotions like fear, anger, or happiness in the first place. Still, even if there’s literally nobody home, what does constantly being cruel to something that convincingly presents itself as sentient do to you as a person?

CodeWall raised a $4M Pre-Seed for Offensively Good Security.

Zenzap closed a $33M Series A for Team Chats With AI Agents.

North.cloud landed a $60M Series B to Unify Speed Intelligence.

Databricks (Bellevue, WA): Software Engineering Intern, Summer 2027

NVIDIA (Austin, TX): ASIC Verifications, SoC’s and GPU’s, Entry-Level

Anthropic (San Francisco, CA): Risk & Insurance Specialist, Mid-Level

OpenAI (San Francisco, CA): Audio Engineer, Live Video, Senior-Level