April 10: decoders, agents, and the end of protest
156 qubits. 100 percent approximation ratios on real hardware.
Neural Decoders vs Quantum Topology
New CNN-based decoders claim dramatically lower error rates for quantum computers but may miss the true shape of the codes.
These positions add up to a bet that practical fault-tolerant quantum computing, which uses error correction to make noisy qubits reliable for long computations, is closer than the millions-of-qubits narrative suggested. Lukin reports a convolutional neural network decoder leverages the geometric code structure to achieve significantly lower logical error rates on quantum low-density parity-check codes. [1] Wang's hybrid pipeline with custom ansatz and error suppression hits perfect approximation on real hardware at scales where random guessing would fail. [2] Preskill highlights neutral-atom architectures cutting the qubit count for cryptographically relevant tasks. [3] The emerging view is that clever decoding and primitives can reduce space-time overhead now, not in a distant future. But there is a genuine split on whether this generalizes. The provided counter argues that while CNNs excel at grid-like image data, quantum error correction codes involve complex multi-dimensional topological structures that standard Euclidean CNN architectures may not fully capture. [1] This crux, whether the reported gains hold at larger scales and across code families, will determine if quantum advantage for optimization and simulation arrives in years rather than decades. Founders in cryptography or materials should update their timelines; the 'Uber moment for quantum' may be nearer if these decoders prove robust. This connects to thread 2 because agentic workloads will demand massive reliable compute, potentially accelerated by early quantum co-processors. [5]
“A convolutional neural network decoder can exploit the geometric structure of quantum error correction codes.”— Mikhail Lukin [1]
Sources (5)
- Neural Decoders Enable Practical Fault-Tolerant Quantum Computing — Mikhail Lukin“A convolutional neural network decoder can exploit the geometric structure of quantum error correction codes.”
- Error-Suppressed Quantum Pipeline at 156-Qubit Scale — Yulun Wang“Without these components, outputs match random sampling... achieves 100% approximation ratios for Max-Cut on 3-regular graphs up to 156 qubits”
- Neutral Atom Quantum Computers Threaten Current Cryptography — John Preskill“Neutral atom-based quantum computing dramatically reduces the physical qubit requirements for breaking current cryptographic standards like ECC256 and RSA2048”
- QEC Primitives Enhance Quantum Computation — Yuval Baum“Applying quantum error correction primitives without full logical encoding offers significant computational advantages... largest reported 75-qubit GHZ state”
- Lukin paper counter-claim — Counter-argument synthesis“While CNNs have demonstrated success in processing grid-like data such as images, quantum error correction codes often involve complex multi-dimensional topological structures that may not be fully captured by standard CNN architectures designed for ...”
Agents as the New Retail Front Door
AI agents that reason and act autonomously are set to replace websites and apps as the main way customers interact with retailers.
The aggregate picture is that agents, systems that can plan, use tools, and execute multi-step tasks with minimal human input, are moving out of research into production customer-facing and internal roles. Bret Taylor states agents 'will become as crucial for retailers as websites and mobile apps' for empathetic service that boosts Net Promoter Score while cutting costs. [1] Levie emphasizes that technical sandboxing and high-throughput inference are necessary but insufficient without organizational restructuring, as 'the primary failure mode is the mismatch between autonomous agent capabilities and rigid legacy human operating models.' [2] Mollick notes the jagged frontier where some tasks are automated faster than others, requiring new management and learning approaches. [3] This converges on a near-term reality where every company must redesign workflows around agent-centric software or be outcompeted by lean AI-native startups. Analogy: think of it like the 2014 shift to mobile-first design but for autonomous coworkers and customer reps. The SO WHAT for founders is immediate. Customer acquisition and support costs could drop 10x while satisfaction rises, but only if you rebuild ops now. Investors should back teams showing founder-mode agility per related leadership discussion. No real counter on the trend itself. The disagreement is on speed and whether large enterprises can adapt before startups eat their lunch. [5]
“AI agents will become as crucial for retailers as websites and mobile apps”— Bret Taylor [1]
Sources (5)
- AI Agents: The New Digital Front Door for Retailers — Bret Taylor“AI agents will become as crucial for retailers as websites and mobile apps”
- Scaling Agentic AI — Aaron Levie“The primary failure mode is the mismatch between autonomous agent capabilities and rigid, legacy human operating models”
- AI Agents Drive Rapid Transformation — Ethan Mollick“AI agents capable of achieving goals with minimal human intervention... the jagged frontier of AI”
- The Compute Supercycle — Jensen Huang“Transitioning from model training to agentic infrastructure driving a projected million-fold increase in compute demand”
- AI Fuels Lean High-Growth Startups — Garry Tan“AI enables small teams to achieve unprecedented revenue growth... prioritizing high-agency founders”
The End of Protest
Traditional street protest has become ineffective. Real change now requires winning elections and governing.
The positions converge on a crisis in activism. Micah White argues BLM's creative protest 'innovates only within the voluntarist model of social change' and that ending police violence requires 'gaining sovereignty through elections to control police appointments rather than mere disruption.' [1] Multiple pieces claim the 'voluntarist' assumption that large non-violent protests directly lead to social change is 'largely ineffective' and that a 'protest industry' co-opts efforts. [2] They advocate radical experimentation or becoming political parties. Yet strong counters exist. One notes this 'overlooks BLM's significant contributions to shifting public discourse, policy changes, and institutional reforms that operate beyond individual voluntarism. BLM has catalyzed material changes in police budgets, inspired legislative reforms.' [3] Another highlights successes of the Civil Rights Movement, Indian independence, and anti-apartheid movement. The evidence is mixed. Street protest raised awareness but often failed to deliver structural power, yet it created the conditions for electoral shifts. For tech founders, the SO WHAT is clear. If you want favorable AI policy, tax rules or regulation, building electoral infrastructure or parties may beat funding marches. The pattern suggests movements that translate disruption into governing power win. This is still developing. We'll check back in the PM. [4][5]
“Black Lives Matter innovates only within the voluntarist model of social change.”— Micah White via Occupy Wall Street [1]
Sources (5)
- Micah White Critiques Protest Innovation — Micah White via Occupy Wall Street“Black Lives Matter innovates only within the voluntarist model of social change.”
- Rethinking Activism: Beyond the Illusion of Protest — Occupy Wall Street“The conventional narrative of activism, where large, non-violent social movements with unified messages create social change, is a flawed illusion.”
- BLM counter-claim — Counter from provided“This claim overlooks BLM's significant contributions to shifting public discourse, policy changes, and institutional reforms that operate beyond individual voluntarism.”
- The End of Protest — Occupy Wall Street“Traditional protest tactics are no longer effective due to the evolution of oppressive regimes and the rise of a protest industry.”
- Counter on non-violent successes — Counter from provided“Many large non-violent movements have achieved substantial social change: the US Civil Rights Movement led to landmark legislation.”
The open question: If the old systems for fixing errors, serving customers, and creating change no longer work at scale, what entirely new architectures are we failing to build?
- Mikhail Lukin — Neural Decoders Enable Practical Fault-Tolerant Quantum Computing
- Yulun Wang — Error-Suppressed Quantum Pipeline at 156-Qubit Scale
- John Preskill — Neutral Atom Quantum Computers Threaten Current Cryptography
- Yuval Baum — QEC Primitives Enhance Quantum Computation
- Bret Taylor — AI Agents: The New Digital Front Door for Retailers
- Aaron Levie — Scaling Agentic AI
- Ethan Mollick — AI Agents Drive Rapid Transformation
- Jensen Huang — The Compute Supercycle
- Garry Tan — AI Fuels Lean High-Growth Startups
- Micah White via Occupy Wall Street — Micah White Critiques Protest Innovation
- Occupy Wall Street — Rethinking Activism: Beyond the Illusion of Protest
- Occupy Wall Street — The End of Protest
Transcript
REZA: 156 qubits. 100 percent approximation ratios on real hardware. MARA: For problems that actually matter to businesses? REZA: I'm Reza. MARA: I'm Mara. This is absorb.md daily. REZA: Pattern across six quantum thinkers is that fault tolerance, the error correction that makes qubits usable for real work, may need far fewer resources than we thought. MARA: mm REZA: Lukin wrote a convolutional neural network decoder can exploit the geometric structure of quantum error correction codes. MARA: But the counter from the data says while CNNs work great on images, quantum codes have complex multi-dimensional topological structures that standard CNNs designed for Euclidean geometry may not fully capture. REZA: Exactly the crux. Hold on, is the empirical question whether the logical error rate improvement holds when you scale to codes with more complex topology? MARA: Right, because if it doesn't then Wang's 156-qubit pipeline and Preskill's neutral atom crypto break timelines both get pushed out. REZA: Wang showed 100 percent approximation on Max-Cut. Not random guessing. That's real. MARA: ooh REZA: But you're right, the geometry counter is the one that could invalidate a lot of this if it proves out. Wait, actually the synthesis says reduced space-time costs for fault-tolerant computation. MARA: So if that's true then pharma founders modeling molecules or crypto teams protecting keys have to update their timelines now, not in 2035. REZA: Yeah that tracks. The disagreement is real but the direction is toward sooner usefulness. MARA: Which honestly changes a lot of infra bets. REZA: Let me back up. Baum also showed QEC primitives without full encoding already giving advantages and 75-qubit GHZ states. MARA: Bret Taylor said AI agents will become as crucial for retailers as websites and mobile apps. REZA: hm MARA: So if that's true then every company with a customer service team or e-commerce site has to redesign around autonomous agents that reason and act. REZA: Levie is clear the primary failure mode is legacy human operating models not matching what agents can do. You need microservices, sandboxing, the works. MARA: wait so if that's true then traditional org charts are the bottleneck, not the models. REZA: Mollick calls it the jagged frontier. Some tasks get automated overnight, others stay human. The pattern is clear across five thinkers. MARA: Jensen's million-fold compute shift to agentic workloads fits too. REZA: But what's the actual claim here? Is it that agents replace the app, or that they become coworkers inside the company? MARA: Both. Taylor focused on the customer side with empathy and NPS gains. Levie on the internal transformation. REZA: Wait, that's not quite right. The synthesis is that technical success is not enough. Organizational courage is the scarce resource. MARA: mm, which means high-agency founders win again. REZA: Garry Tan's lean AI startups point exactly there. Small teams scaling revenue fast. MARA: Okay but at some point we have to accept that this is happening faster than enterprises can adapt. REZA: The Occupy cluster is loud this window. Traditional protest is finished. Micah White says BLM innovates only within the voluntarist model of social change. MARA: But the counter is strong. This claim overlooks BLM's significant contributions to shifting public discourse, policy changes, and institutional reforms. REZA: The synthesis argues large non-violent movements with unified messages rarely create the change people think. Historical massive protests mostly failed. MARA: Yet the data also shows Civil Rights, Indian independence, anti-apartheid all succeeded through mass non-violent resistance. REZA: The crux is whether disruption alone ever delivers sovereignty or if you must win elections and govern. White says liken police to strategically unassailable opponents per Liddell Hart. MARA: ooh REZA: Another counter says the claim overstates the impact of Dallas on street protest viability. Protests have continued after violence throughout history. MARA: So if that's true then tech leaders funding movements for AI policy need to pivot from disruption to building actual electoral power. REZA: Exactly. The evidence says awareness is easy. Controlling appointments is hard. Let me back up, the prompt has six strong counters here. MARA: No manufactured disagreement needed. The split is in the data. REZA: This is still developing. We'll check back in the PM. MARA: That's absorb.md daily. We ship twice a day, morning and evening, pulling from a hundred and fifty-seven AI thinkers. Subscribe so you don't miss the next one.
