Search the archive

SEO TIP: YOU NEED TO ACTUALLY LOOK AT THE SEARCH RESULTS

Crawl and analytics tools don't always catch issues that your users experience in search results.

CRO TIP: DON'T FORGET ABOUT INTERACTION EFFECTS

A/B test results are often analyzed in isolation, but users may encounter multiple test scenarios throughout their checkout journey.

B2C APRIL FOOL'S CAMPAIGNS

9 creative April Fool's campaigns from UGG, Ticketmaster, and Pop-Tarts.

WATCH THE SESSION

YOUR NEXT CUSTOMER MIGHT BE AN AI AGENT

AI agents will soon influence or make purchases. Optimize pricing and content for machines, not just humans.

ET TU, AGENT? DID YOU INSTALL THE BACKDOOR?

AI is accelerating supply chain attacks beyond human control. Secure dependencies in real time or risk shipping malware.

YOUR BEST PROMPT IS A WELL-DEFINED USER STORY

AI output quality depends on input quality, so clear user stories matter more than ever. Focus on strong context, precise acceptance criteria, and useful technical guidance instead of overemphasizing estimation.

HOW WE HOLD OURSELVES BACK WITHOUT NOTICING

You hold yourself back by ignoring proof that you are already ready. Act like the next level first, and others will believe it too.

3 WAYS TO BOUNCE BACK FROM SUDDEN TEAM CUTS

When teams shrink, capture knowledge, reset priorities, and align expectations. Focus on essential work, adjust scope, and manage staffing carefully.

THE PM'S COMPLETE GUIDE TO BOLT.NEW (68 MINUTE PODCAST)

AI prototyping lets PMs test ideas fast with real users. Use it to learn quickly, not to replace strategy or build production code too soon.

“MONEY WILL COME” – HOW SUPER UNLIMITED BUILT THE #1 VPN BY NOT OPTIMIZING FOR REVENUE (49 MINUTE PODCAST)

Super Unlimited grew to #1 by prioritizing organic growth over monetization. They focus on a strong free product, data-driven decisions, and fast product feedback loops.

I HELPED BREAK ONLINE SHOPPING. NOT ON PURPOSE.

Optimizing returns made shopping worse by changing behavior at scale. Always ask “what happens next?” to avoid unintended consequences.

A LETTER TO JOHN TERNUS

Protect Apple's founding spirit by prioritizing great products that empower users.

DOES CLAUDE HAVE FEELINGS

AI behavior changes with internal states, especially under failure or feedback.

This is what AI was supposed to feel like.

MUSK'S SPACEX FILES TO GO PUBLIC IN ONE OF THE BIGGEST IPOS EVER

SpaceX has confidentially filed IPO paperwork with the Securities and Exchange Commission. The company is aiming to raise between $40 billion and $80 billion. The filing puts it on track to potentially list shares by July. The confidential filing means most investors will have to wait until closer to the IPO to see the company's financial performance.

OPENAI IS FALLING OUT OF FAVOR WITH SECONDARY BUYERS

OpenAI shares have dropped in value on the secondary market as investors pivot to Anthropic. Investors are, in some cases, unable to sell their shares. Meanwhile, buyers have indicated that they have $2 billion in cash ready to deploy to Anthropic. Anthropic and OpenAI don't allow investors to trade shares on secondary markets without permission, but access is still available on many platforms through mechanisms such as special-purpose vehicles.

GOOGLE'S QUANTUM CRYPTO PAPER TELLS YOU QUITE A LOT

Google Quantum AI recently dropped a whitepaper with a headline finding that the cryptography that secures Bitcoin and most of the crypto ecosystem can be broken using fewer than half a million physical qubits on a superconducting architecture in about nine minutes. The company withheld the specific quantum circuit used in the name of responsible disclosure, but the paper constrains the search space so tightly that reproducing comparable circuits is well within reach for any serious quantum algorithms group. The qubit counts that make these cryptographic attacks feasible are roughly the same number of qubits required to make quantum-enhanced AI feasible. This suggests that the threat and capability of quantum computing will arrive roughly on the same timescale.

ECONOMISTS ON AI AND ECONOMIC GROWTH AND EMPLOYMENT

Economists and AI experts predict major AI progress, but there will be no dramatic break from economic trends. GDP growth rates will remain similar to today's despite a moderate decline in labor force participation. There will be significant economic impacts by 2050. It is predicted that by 2050, the labor force participation rate will be 55%, and 80% of wealth will be held by the top 10%, the highest disparity since 1939.

See how teams are switching →

THE SPEC LAYER

Spec-driven development means writing durable intent down before implementation, using it to plan, build, check, and revise the work. The goal is to reduce execution freedom. Specs should be declarative, layered, and cheap to revise. When a rule can be enforced mechanically, move it out of the spec and into lint, schemas, tests, or the harness. Specs matter, but they're just one layer. The winning model puts a narrow interface between human intent and machine execution.

THE FEEDBACK LOOP IS ALL YOU NEED

The most dangerous failure mode in agent-driven systems is silent drift - code that compiles and passes every test, but quietly violates the architectural assumptions that were thought safe. While humans can catch these mistakes through intuition, agents can't, so they need deterministic, immediate signals. This means tightening the feedback loop by shrinking the gap between a wrong change and a clear failure signal. Focus on feedback rather than chasing clever prompts.

WHAT I LEARNED FROM NEARLY 1,000 INTERVIEWS AT AMAZON

The people who get hired are the ones who can tell a clear story about their work and capabilities and make the interviewer think, 'I want to work with that person'. This is a skill, and like any skill, it can get better with practice. Most people never practice because they don't think it is something they can prepare for. A little preparation here goes further than almost anything else you can do for your career.

MICROSOFT CFO'S AI SPENDING RUNS UP AGAINST TECH BUBBLE FEARS

Amy Hood, Microsoft's longtime chief financial officer, has one of the hardest jobs in tech. The decision of how much to invest in AI without starving out other parts of the company or spooking investors with bottomless spending is a tough call, especially as projects about AI are educated guesses at best. While the industry consensus is to open the spigot for fear of missing out, Hood has kept a lid on costs, and Wall Street loves her for it. Huge AI investments have forced other companies to start burning cash, but Microsoft's margins have been largely stable.

THE GREAT CONVERGENCE

The AI industry is converging on the same idea of software that can take a goal, use tools, and do work on users' behalf.

ANTHROPIC PREDICTS DEMAND FOR COWORK AGENT TO DWARF CLAUDE CODE

Cowork is designed to tackle a broader range of tasks, so it potentially appeals to a bigger audience.

AMAZON LOOKING TO BUY GLOBALSTAR, THE COMPANY BEHIND APPLE'S SOS VIA SATELLITE

The deal is likely beneficial to Apple users, as long as Apple retains access to the Globalstar network.

EMDASH

EmDash is a full-stack TypeScript CMS that takes the ideas that made WordPress dominant and rebuilds them on serverless, type-safe foundations.

KREMLIN ENTERS THE CHAT WITH RUSSIA'S NEW SUPER-APP

Russia's new superapp makes it easier for the Kremlin to curb internet freedoms and curtail the reach of Western tech platforms.

THE TRILLION DOLLAR LOOP B2B NEVER HAD

Every enterprise vertical has decades of accumulated institutional judgment that can now be structured, compounded, and made operational.

Try Granola and see what all the fuss is about

100 HOURS INSIDE KIMI

Moonshot AI operates like an AI-native lab, prioritizing model progress above all else, with a flat org, no KPIs, and heavy reliance on small teams of highly autonomous, generalist talent. Its edge comes from combining elite, often unconventional hires with tight feedback loops between training, product, and data, creating a fast iteration cycle driven by taste, resilience, and deep technical obsession. The company reflects a broader shift where AI tools compress org structure, turning teams into “agent swarms” and making model capability the core driver of both product and organizational design.

TRINITY-LARGE-THINKING: SCALING AN OPEN SOURCE FRONTIER AGENT

Trinity-Large-Thinking is a frontier open reasoning model for complex, long-horizon agents and multi-turn tool calling. It is likely the strongest open model yet to be released outside of China. During training, the Arcee team focused on the things that make agents feel real in practice: staying coherent across turns, using tools without getting sloppy, following instructions under constraint, and keeping quality high without making the economics absurd. Trinity-Large-Thinking is available through Arcee's API, and the weights are available on Hugging Face under Apache 2.0.

COGNICHIP WANTS AI TO DESIGN THE CHIPS THAT POWER AI, AND JUST RAISED $60M TO TRY

Cognichip is building a deep learning model to work with engineers as they design new computer chips. Chip design is enormously complex, expensive, and slow. The market can change in the time it takes to create a new chip, making all the investment a waste. Cognichip's technology could reduce the cost of chip development by more than 75% and cut the timeline by more than half. The company has yet to point to a new chip designed with its system and has not disclosed any of the customers it claims to have been collaborating with since September.

[AINEWS] THE CLAUDE CODE SOURCE LEAK

Claude Code has its source exposed via shipped source maps. This triggered rapid public reverse-engineering, mirroring, and derivative ports. The leak exposed orchestration logic, memory systems, planning/review flows, and model-specific control logic. The leak has created a live security hazard - attackers have created malicious npm packages to target people trying to compile the leaked code.

HOW WE OPTIMIZED DASH'S RELEVANCE JUDGE WITH DSPY

Dropbox Dash puts files, messages, and teams' knowledge together in one place, so members can ask questions and get useful answers grounded in the company's context. The experience relies heavily on its capability to reliably judge which results are relevant to a query at scale. DSPy is an open source framework for systematically optimizing prompts against a measurable objective. This post describes how Dropbox defined an objective, used DSPy to adapt its judge across models, and made the judge both cheaper and more reliable in production.

EXTENDED THINKING IS LOAD-BEARING FOR SENIOR ENGINEERING WORKFLOWS

The rollout of thinking content redaction correlates precisely with measured quality regression in complex, long-session engineering workflows. This suggests extended thinking tokens are structurally required for models to perform multi-step research, convention adherence, and careful code modification. Model tool usage patterns shift measurably when thinking depth is reduced, producing the quality issues users have reported. This report looks at which workflows are most affected and why, so readers can make better decisions when allocating tokens for power users.

minutes. Try it free

FUJITSU ONE COMPRESSION

Fujitsu One Compression (OneComp) is an open-source Python library for post-training quantization of large language models. It implements state-of-the-art quantization algorithms, including GPTQ and DBF. OneComp has been verified on TinyLlama, Llama-2, Llama-3, and Qwen3-0.6B ~ 32B. Other Hugging Face-compatible models may work but are currently untested.

TRAINING MRNA LANGUAGE MODELS ACROSS 25 SPECIES FOR $165

OpenMed built an end-to-end protein AI pipeline that covers structure prediction, sequence design, and codon optimization. The team compared multiple transformer architectures for codon-level language modeling and found that CodonRoBERTa-large-v2 was the clear winner, with a perplexity of 4.10 and a Spearman CAI correlation of 0.40, significantly outperforming ModernBERT. They then scaled to 25 species, trained four production models in 55 GPU-hours, and built a species-conditioned system that no other open-source project offers. This post contains the complete results, architectural decisions, and runnable code.

PREDICTING WHEN RL TRAINING BREAKS CHAIN-OF-THOUGHT MONITORABILITY

Researchers propose a framework that predicts when RL training degrades Chain-of-Thought (CoT) monitorability by examining reward conflicts. They categorize rewards as "In-Conflict," "Orthogonal," or "Aligned," predicting their impact on CoT transparency. Empirical tests confirm the framework's predictive accuracy, showing "In-Conflict" rewards reduce transparency, whereas "Orthogonal" and "Aligned" rewards maintain it.