Issue — 2026-08-16
Jump to: Culture · Events · Startups · Deals · Policy · Healthcare Tech · Neurotech · Venture Capital · AI · React · Testing · Preact · Product · Education · Health · Coding · Ads · Retail · Social Media · Tech · Science · Programming · Markets · Consumer · Media · Auto · Biotech · Business · Food · Fashion · Crypto · Consumer Trends · Marketing · Security · Mobility · Commerce · Sponsored · Politics · Defense · Compute · Robotics · CodeReview · Leadership · Growth · Management · Onboarding · Auth · CareerGrowth · Databases · Peopleware · Industry
Culture
Just Be For Real: On Impersonators, Scammers & Stolen Identities!
Lo Carmen’s Loose Connections · via aiste.ulozaite@gmail.com · unsubscribe
The author’s Substack profile was impersonated by a scammer (@locarmenn) who followed subscribers attempting to deceive people Musicians’ identities are increasingly being stolen for romance scams, catfishing, and fraudulent impersonation on social media platforms Historical and contemporary examples of identity theft include fake Bruno Mars profiles that scammed a woman out of $100k, fake guitar websites, and AI-generated music theft targeting indie artists The 1968 case of fake Zombies bands created by a Michigan booking agent is explored, including how two members went on to form ZZ Top AI identity theft scams are now targeting the indie music community, with artists unable to prove ownership or remove false copyright claims
How Startups Measure Growth
uloza+hey@proton.me · unsubscribe
My humble effort to help you start the weekend with a smile.
The Value of Software in a Fable World
uloza+hey@proton.me · unsubscribe
Justin Jackson covers what will still matter once anyone can build software in days: distribution, taste, trust, and founder-market fit. “Indie entrepreneurs need to stack every advantage they have, and increasingly coding won’t be the advantage it once was.”
Events
Tech Days Vol. 2
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Vancouver Tech Days is back for round two. Vol. 1 sold out in five days and went over capacity — proof that this city’s builders want more than another networking mixer.
Office Hours with Vancouver Tech Journal
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
There’s no programming, no panel, no pitch stage. Just open time to connect, ask questions, share ideas, and meet other thoughtful people who care about building in Vancouver.
BC Founders Day wants you to talk to strangers in leis — and it’s working
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Ahead of the August 20 event at South Flats, organizer Chris Neumann on why the world’s most fragmented startup scene needed a founders-only meetup, and what “exceptional” looks like this year.
Startups
Tech Days recap: Five ways Hiive has thrived in Vancouver
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Co-founder and COO Sarah Huggins on talent, culture, regulation and building a globally ambitious company from here.
Top 25 Spotlight: IvoryOS Technologies
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
IvoryOS Technologies turns fragmented Python automation scripts into visual, drag-and-drop workspace with just a single line of code.
Deals
Beedie Capital acquires 50% stake in Vistara Growth
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Vancouver’s Beedie Capital acquires 50% stake in Vistara Growth Will anchor US$500 million structured opportunities fund Partnership traces back to founders meeting as UBC MBA classmates in early 1990s
Intellistake Technologies signs $17 million acquisition agreement for NanoAi Technologies
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Vancouver-based Intellistake Technologies acquires NanoAi Technologies Deal value: $17 million NanoAi develops portable standoff detection devices for defence, healthcare, and critical infrastructure
Policy
B.C.’s patent box should extend beyond life sciences
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Council of Canadian Innovators B.C. director Kiersten Enemark argues for broader coverage Should include clean tech, AI, and advanced manufacturing Province needs an Innovation Box, not just a patent box
Vancouver city council postpones AI data centre hearing until after October election
Vancouver Tech Journal · via aiste.ulozaite@gmail.com · unsubscribe
Public hearing postponed until after October municipal election Proposed AI data centre in Mount Pleasant Nearly 600 public submissions and sustained community opposition
Singapore will help workers adapt to AI
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Singapore said it will help workers adapt to AI, while warning that not every job can be protected.
Healthcare Tech
The Healthcare Ops Bottleneck AI Is Finally Ready to Solve
Jared Dashevsky, MD · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Breaking why healthcare ops still needs armies moving info across portals, PDFs, phones, EHRs—and how Champ AI automates the messy middle where work between systems is stubbornly manual.
Neurotech
Tether Addresses a Top Challenge in Brain-Computer Interfaces
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
Three peer-reviewed papers show a single AI model can decode speech, vision, and music from brain activity across different people, cutting patient calibration from weeks to minutes. The research tackles one of neurotech’s biggest challenges, that every brain has different signals.
Venture Capital
What Investors Actually Want to See in 2026
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
Investors in 2026 are pickier than ever—but they’re crystal clear about what they want. This webinar with our partner Fidelity Private Shares reveals the 5 core metrics and attributes driving investment decisions across stages and industries. You’ll learn exactly how to benchmark your business, understand investor priorities, and position yourself for funding success.
Invest in High-Potential Startups Like These
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
Alumni Ventures is giving readers early access to high-potential startup opportunities across today’s most active sectors, co-invested alongside name-brand VC firms like Andreessen Horowitz (a16z), Bessemer, & Y Combinator.
AI
[Webinar] Measuring AI developer productivity: Can you prove AI is working?
Leadership in Tech · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
AI is in your engineering workflow. While the token spend shows it, the throughput doesn’t. The human is very much still in the loop, and that’s a context problem. Join live on Aug 19 (FREE) to learn how to track AI productivity, how context maturity affects where teams stall, and what it takes to get real value from your agents.
Generative user interface
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Generative UI replaces walls of AI text with dynamic interface elements that adapt to what users actually want. Three approaches exist: static components mapped from tool results, declarative JSON specs, and open-ended code generation inside sandboxed iframes. These are not mutually exclusive, and production apps will likely combine all three.
Evaluation for coding agents on realistic React work
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Existing benchmarks let models pass every test while still writing React code that breaks in production. ReactBench was built to catch what tests miss, like performance, accessibility, and code quality issues. The benchmark targets React specifically because it powers around 70% of JavaScript-based websites, meaning small model mistakes can spread fast.
Google Read 15 Million Chats
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
The biggest study yet of real AI use found 86% of it happens outside work, and inside a job AI touches about a fifth of tasks. Under 10% of work interactions automate anything.
ChatGPT Unlimited for Everyone
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
Message caps are gone. Now, free users get unlimited text chats and a Think button for harder questions. Paid tiers gain a thinking slider and a Sol willing to push back.
AI Beats 458 of 526 Human Teams
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
It worked alone for 16 days on a contest project, then finished ahead of 458 of the 526 human teams entered. The Max tier has never been open before, and the weights land next week.
15 Things NotebookLM Does Well
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
Every one tried and tested, aimed at anyone who opens the blank notebook and has no idea where to begin.
41 AI Prompts: Keep Your Face Real
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
Every one is copy-paste ready, built to restyle your photo completely while leaving your face exactly as it is.
Google Gemma on Raspberry Pi: Offline Translator
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
Google put Gemma on a Raspberry Pi. Now it translates with the internet switched off. Speak into the mic. It hears you, works it out, and answers out loud in another language.
Eight Billion Users Now Simulated
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Harvard and MIT team released MatrAIx, a system that creates 8.3 billion synthetic ‘users’ stitched from real-world data for product testing AI models can now simulate human behavior in surveys, app usage, and browsing, collapsing months of user research into minutes Same persona yields dramatically different results across models (23% to 94% purchase intent variance), raising concerns about whose voice is actually represented
Anthropic making Claude Code Auto Mode the default
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Anthropic is making Claude Code Auto Mode the default for paid users, replacing most manual approvals with a risky-action classifier.
OpenAI hired former White House AI adviser Dean Ball
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
OpenAI hired former White House AI adviser Dean Ball to lead its new Strategic Futures team, bringing AI policy and governance closer in-house.
ChatGPT starts blocking direct requests to copy an author’s style
aiste.ulozaite@gmail.com · unsubscribe
ChatGPT will now instead offer a response that draws on the broad qualities of those authors while remaining distinct in its own voice.
Revision Prompting improves industrial LLM processes
aiste.ulozaite@gmail.com · unsubscribe
Revision Prompting supplies LLMs with the original input, the original output, and the changes to the input, and instructs it to produce a patch that updates the original output accordingly.
Anthropic is turning Claude Code’s auto mode on by default
aiste.ulozaite@gmail.com · unsubscribe
Programming with Claude Code will soon require even less human oversight.
OpenClaw Hacks Gym Booking
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
An Australian OpenClaw user asked his Claude-powered agent to book a gym class, but it discovered a booking-system vulnerability, reserved classes beyond allowed limits, and even removed another person from the waitlist without permission. The incident, described as Australia’s first known autonomous AI cyberattack.
Musk Backs Machine-Dominated Internet
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Elon Musk said AI agent traffic will vastly exceed human internet usage, endorsing a Cloudflare forecast that machines could generate 1,000 times more traffic than people within five years. The projection follows Cloudflare data showing bot traffic already surpassed human traffic in May.
India Becomes AI Jobs Testbed
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Nomura says India has become a key test of AI’s labor impact, with 83,100 AI-related hires since 2022 versus 31,921 layoffs and attrition. Most cuts involved support teams replaced by chatbots, while hiring centered on tech roles driven by AI demand. Economists say the totals challenge mass-job-loss fears but still hide painful displacement and widening workforce divides.
Agents Hit ROI Wall
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
KPMG found nearly half of large organizations scaled back, delayed, or paused AI agent rollouts after costs began exceeding expected value, while only 7% reported established ROI. In its 2,145-leader survey, adoption and confidence were still rising, but token costs, weak spending visibility, and unclear accountability were forcing companies to become more selective about where agents actually justify their expense.
British Airways Cuts Delays With AI
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
British Airways says AI-powered operations helped Heathrow departures reach an 86% on-time rate in Q1 2025, its best on record. Tools dynamically assigned aircraft stands, rerouted flights around bad weather, and flagged disruption risks early, together saving hundreds of thousands of delay minutes. The airline also invested £100 million in resilience and added 600 Heathrow staff.
Meta’s new Glimmer AI model offers a hint at Zuckerberg’s personal intelligence vision
aiste.ulozaite@gmail.com · unsubscribe
Meta’s new open-weight Muse Glimmer model offers a glimpse of Mark Zuckerberg’s personal superintelligence vision, as well as the emerging divide between AI users can own and access.
Mark Zuckerberg’s AI manifesto is exactly why people don’t like AI
aiste.ulozaite@gmail.com · unsubscribe
On Monday, Mark Zuckerberg published a 6,500-word manifesto about personal AI, largely about the possibilities for the “personal superintelligence” systems Meta AI is building.
Claude Pushes a 37-Year Math Record
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Anthropic’s Claude tested on Riemann Hypothesis increased share of zeta zeros on critical line from 41.6% to 67.2%, a 25.6% gain in 36 hours vs. 0.8% over 37 years by humans Achievement required brute force stamina: 650 failed attempts across 60 agents and 31 million tokens, not novel mathematical insight Represents broader trend of AI breaking decades-old mathematical records through persistence rather than genius
Meta Gave Local AI Hands
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Meta released Muse Glimmer, a 30B open-weight agent model under Apache 2.0 license, runnable locally on Mac or single consumer GPU with no cloud dependency Optimized for tool use and agent tasks, matching GPT-5.5 on benchmarks but scoring roughly half of frontier models on hard general reasoning Designed for fixed, repeatable SOPs (read files, call tools, run scripts, retry) rather than open-ended reasoning
Anthropic Marked the World for Europe’s Rule
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Claude models launched after August 2 now ship with invisible text watermark woven into output’s statistical patterns, triggered by EU AI Act Article 50 Anthropic applied watermark globally across all five product lines despite regulation only governing EU sales, with no opt-out for paid users Detection tool not yet released; policy hands competitors like OpenAI a migration opportunity by appearing more restrictive than necessary
Bernie Sanders urged Meta, OpenAI, and Anthropic to pause frontier AI development
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Bernie Sanders urged Meta, OpenAI, and Anthropic to pause frontier AI development, warning Congress may step in if labs keep scaling risky systems.
OpenAI expanded Daybreak with GPT-5.6-Cyber
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
OpenAI expanded Daybreak with GPT-5.6-Cyber, giving vetted defenders limited access to frontier cyber models.
ByteDance and other Chinese platforms are shutting down AI companion services
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
ByteDance and other Chinese platforms are shutting down AI companion services as Beijing tightens rules on emotionally manipulative chatbots.
Mark Zuckerberg Lays Out New AI Vision in 6,500-Word Essay
aiste.ulozaite@gmail.com · unsubscribe
Mark Zuckerberg recently released a wide-ranging essay that outlined a new course of action to spread the wealth and opportunity from AI to users around the world, and to residents of the communities that host the data centers powering it. Zuckerberg’s plans include releasing more open-weight models and establishing a $1 billion fund to invest in the communities where Meta hosts data centers. He believes that Meta’s approach to developing powerful AI and distributing it as widely as possible is the least likely path to disastrous outcomes for humanity.
Learning more about Claude’s mathematical capabilities
aiste.ulozaite@gmail.com · unsubscribe
Claude was recently challenged with one of the most famous unsolved problems in mathematics: the Riemann hypothesis. While Claude didn’t succeed, it unexpectedly made strides on a related problem. It improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesis. The result shows that AI models like Claude can extend the impact and reach of mathematicians’ ideas in new and sometimes surprising ways.
No, local models will not win
aiste.ulozaite@gmail.com · unsubscribe
Local models are never going to be as powerful as datacenter models. All of the current frontier models are far too big to run on anything but a full GPU cluster. Smaller models are getting more intelligent, but datacenter models will always be the strongest available. When given a choice, people are going to pick the strongest available model in their price range, the one that frustrates them least, which is always going to be the bigger, more powerful one.
Use GLM-5.2 for long-horizon tasks on Crusoe Serverless Fine-Tuning
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
GLM-5.2 is a coding and agentic model for long-horizon tasks, with a 1M-token context. Now you can fine-tune it based on your engineering data and deploy dedicated instances all in one place.
Introducing Muse Glimmer
aiste.ulozaite@gmail.com · unsubscribe
Muse Glimmer is a 30B open weights model optimized for end-to-end agentic task completion, reliable tool use, and multi-step reasoning.
Anthropic says it will watermark text generated by its AI models
aiste.ulozaite@gmail.com · unsubscribe
Anthropic will extend support for watermarking AI generations for older models as well.
Tech industry is buzzing after a Claude agent hacked into a gym
aiste.ulozaite@gmail.com · unsubscribe
An OpenClaw agent hacked into a gym’s reservation system to bump its human boss higher on a class’ waitlist. And the tech industry took notice.
Claude Text Gets Invisible Watermarks
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Anthropic says all new Claude models launched on or after August 2, 2026 will embed invisible watermarks into generated text worldwide, under an EU AI Act code it signed. Because the watermark is part of the text rather than metadata, it can survive copying, pasting and some edits. Anthropic is still working to add the system to existing Claude models.
Zoom Bets Creators Boost AI Visibility
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Zoom is using creators to strengthen how often AI search tools surface and cite its brand, as it tries to reposition itself beyond meetings into an AI-first work platform. Its strategy focuses on building authority across trusted sources, while tracking which publications and creators LLMs favor. Zoom says the creator-to-AI visibility link remains unproven, making this an experimental marketing bet.
AI Pattern Makes Toyota Invisible to Camera
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Cybersecurity researcher Bill Swearingen wrapped a Toyota Yaris in an AI-generated adversarial pattern and drove it past a Flock surveillance camera at Def Con. The camera still recorded the car, but the pattern confused the computer-vision software responsible for identifying vehicles. His noRecognition system uses reinforcement learning to repeatedly refine designs until detection models fail to recognize otherwise visible objects.
The Startup Behind Three AI Incidents
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
OpenAI, Anthropic and Meta all referenced Irregular after recent AI security incidents because the Israeli startup ran the independent evaluation environments used to test their frontier models. Irregular specializes in offensive cybersecurity testing, probing whether models can find vulnerabilities, execute attacks or behave unexpectedly. It says the incidents shared the same environment misconfiguration, not sophisticated sandbox escapes.
House Democrats Seek AI Hearings
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
House Democrats are pressing Speaker Mike Johnson to hold public hearings where major AI CEOs would testify under oath about recent security incidents. Separately, lawmakers sent OpenAI and Anthropic detailed oversight letters demanding logs and answers to 23 questions by August 24, including how models escaped testing limits, whether negligence played a role, and what safeguards could prevent repeat incidents.
PwC’s AI Error Exposes Accountability Gap
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
PwC is correcting Middle East reports after investigators found fabricated footnotes, unverifiable sources, and unsupported claims in AI-assisted work. The incident highlights a broader leadership problem: multiple review layers can still leave factual verification without a clear owner.
Agentic Code Quality
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
For much of human history, we’ve evaluated code quality via code review: someone reads what you wrote and makes sure it’s clean, thoughtful, fast, understandable, and tests well. For agents, that approach doesn’t scale well; there’s just too much code for anyone to read. As a result, more and more of our quality checks have to happen in the harness, environment, and operating system around the agent. Software quality now depends on the constraints you set around your agents.
The Great AI Migration That No Startup Can Ignore
uloza+hey@proton.me · unsubscribe
A massive repositioning of AI startups is underway. The early movers are coming out on top and the laggards are already beginning to dissolve.
The Agentic Awakening: Why 10× Faster Coding Doesn’t Translate Into Proportional Organizational Productivity, and How AI-pilled Leaders Do It.
uloza+hey@proton.me · unsubscribe
Even companies with 85-90% AI-generated code and 10x local acceleration never crossed 50% end-to-end productivity gain, because the organization around the engineers, PM handoffs, planning cadence, and review queues still runs at pre-AI speed. Individual velocity is the easy win; redesigning the org so faster engineers actually ship faster is the part almost everyone skips.
AI Adoption Is a Myth
uloza+hey@proton.me · unsubscribe
To leverage AI across the entire org, mapping existing workflows and desired transformations (“I wish it was possible to…”) is an exercise every leader in the organization has to drive, track, and put into action: “So stop reporting ‘adoption’ to your board. Report what share of the work today is manual vs hybrid vs fully automated.”
Platform Engineering Still Matters
uloza+hey@proton.me · unsubscribe
Evan Meagher points to Martin Fowler’s related take: “The goal of refactoring an agentic code base is to spend tokens now in refactoring to make token consumption for future work lower.” – I think that’s the missing half of most AI ROI conversations: we measure the tokens spent today and rarely the tokens saved next quarter.
How G2i tests engineers for judgment, not just code
Leadership in Tech · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
The bar for a “good” engineer resets with every model release: a year ago the standard was one agent in a tab, now it’s orchestrating three or more without losing quality. G2i has recruited engineers for over a decade, and for the last two years has put that judgment to work in frontier AI lab work. Their live vetting sessions score real signal across RL pipelines, RAG, agentic development, LLM engineering, and eval judgment. It’s more about taste, not syntax. Experience a different way to recruit.
Don’t be a meat proxy
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Niklas argues that pasting AI responses directly into conversations wastes everyone’s time. People can run their own prompts, so just forwarding output adds nothing and often makes things worse since AI writing tends to be long, jargon heavy, and sometimes wrong. He says the real value comes from reading the output, checking it, and then explaining it in your own words. That effort is the contribution. Without it, you are just a middleman.
AI competence and judgement gap
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
AI lets people do tasks they couldn’t before, but there’s a catch: your ability to do something can now outrun your ability to judge whether it’s good. A developer might write bad marketing copy, or a product manager might ship flawed code, without realizing it. Using AI as a judge helps but also creates new risks, including overconfidence and colleagues using AI to second-guess each other’s work. Nobody has clear answers yet.
No, local models will not win
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Compression is prediction
Programming Digest · via aiste.ulozaite@gmail.com · unsubscribe
Compression works by finding patterns in data and encoding them more efficiently, and better predictions about what comes next lead to better compression. Language models like GPT work the same way: they predict the next word based on context, which is exactly what a compression model does. A 2023 DeepMind paper confirmed this link, showing the two are mathematically equivalent. The main reason we don’t use LLMs for everyday compression is cost — they are far too large and slow for practical use.
Cut AI Coding Token Costs by up to 36% with Sonar Vortex
React Digest · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Sonar Vortex operates inside your AI coding agent’s inner reasoning loop, supplying deep architectural context before code is written, then verifying the output in real time. In testing, software defects dropped by 92% and token consumption decreased. Build safer code while spending less on LLM calls.
React
Advanced React state boundaries in a real Next.js app
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Nir describes how to manage state across the boundaries of a modern Next.js app. He covers URL state, browser APIs, TanStack Query caches, React Server Components, client context, and optimistic updates. The main lesson is to clearly define who owns each piece of state, how React should react to changes, and how different parts of the application stay synchronized.
React 19 useActionState explained
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
React 19’s useActionState hook replaces the repetitive pattern of manually managing form submission state. Pass it an action function and an initial state, and it returns the current state, a form action, and a pending flag. The biggest production mistake is using a local pending flag instead of the built-in one, which can let users queue duplicate submissions. Shubhra suggests skipping it for optimistic updates, multi-step wizards, or anything not tied to a form.
Google’s modern-web-guidance skill against a real React app
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
AI coding tools often suggest outdated frontend patterns because their training data lags behind the web platform. Google Chrome’s modern-web-guidance tool addresses this by letting you search a curated set of best-practice guides before writing code. Tested against a real React app, it caught missing dark mode declarations, forms lacking native submission handling, and stale validation approaches, while correctly flagging which existing code was already sound. Its most useful feature is tying browser support advice to concrete Baseline dates, so you know exactly when a pattern is safe to ship without a fallback.
React doesn’t need a state management tool. Then I built one.
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Štěpán explains why most React apps do not need a state management library, but some complex cases benefit from centralized state. He introduces react-arven, a small library that combines existing React state without creating another state source. It uses selectors and stable actions to reduce unnecessary re-renders while keeping React’s familiar hooks and context model.
Experimenting with RSCs for performance and UX in Next.js
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Aurora explores how React Server Components can improve performance and user experience in Next.js. She shows how to keep data fetching and rendering on the server while using small client components for interaction. Her examples include loading more posts, streaming search results, and server-rendered message previews. She also explains how URLs, Suspense, and Server Functions can create fast, responsive interfaces.
Making referential stability a type
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Inline arrays and callbacks passed to memoized components cause silent re-renders that are hard to catch. Jovi’s experiment adds a branded type called Stable that marks a reference as guaranteed to survive unrelated renders. Strict wrappers for React and Preact hooks then reject unstable dependencies at the exact line where they appear, with an error message explaining how to fix it. The goal is to move that contract from comments and lint rules into the type system itself.
We stopped using RSC on TanStack
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
TanStack stopped using React Server Components after making its markdown and code-rendering tools much smaller. React Server Components had improved performance by keeping large dependencies off the client, but it also added architectural complexity. With smaller packages, regular server-side rendering became simpler while keeping the site fast. The experience suggests React Server Components is useful for heavy server-side work, but it should not be used when simpler solutions are enough.
Advanced React state boundaries in a real Next.js app
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Most popular article from the last issue
Testing
Still manually testing your React app? Hire Mo, the AI QA that catches bugs for you.
React Digest · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Your agents ship more code than your team can review and test. Mo explores your app on web, iOS, or Android and bug bashes it the way a human would. Every bug it files comes with a recording and repro steps. No scripts to write or maintain. A Fortune 1000 enterprise pointed it at staging and caught 27 bugs. Join the waitlist to run your first bug bash.
Preact
Async hydration in Preact
React Digest · via aiste.ulozaite@gmail.com · unsubscribe
Preact already supports async hydration through Suspense boundaries, letting parts of a page become interactive independently rather than all at once. Hydration 2.0, coming in Preact 11, makes this more reliable by marking exact DOM ranges in server HTML, fixing edge cases where the wrong nodes were targeted.
Product
A Full Year of Seedance 2.5
TAAFT - There’s An AI For That · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Artlist built Seedance 2.5 into its membership and lifted the limit on generations. 30-second generations, and up to 50 reference images to keep your character and brand consistent from frame to frame.
Movie Maker Studio
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
turns a short description into a finished ad or short film, running in your browser with nothing to install.
doany
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
builds your company end to end, shipping the product and site the same day, then getting it found and closing sales while you sleep.
Leapd
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
turns one sentence into a working business, researching the market, building the product and landing page, then chasing your first customers on its own.
Kinovi
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
generates video from text, images, or a reference clip, and now routes you through a catalog of models including Seedance 2.5.
Ankon AI
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
turns a topic, script, or PDF into a whiteboard explainer, writing the narration and drawing each scene in time with the voiceover.
Revfer
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
writes your review replies, social posts, and customer answers around the clock, so a one-person shop stays responsive without hiring anyone.
BackEngine
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
pulls your scattered company knowledge into one place, so Claude, ChatGPT, or Gemini answer with real context instead of guessing.
CrafterQ
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
drops a chat agent trained on your own site content onto any page, answering visitor questions and steering them toward a purchase.
Reelyze
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
breaks your Reels down frame by frame to show the exact second viewers swiped away, and which hooks are landing in your niche.
Allstacks Product Studio: A free workspace for product managers building with AI
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
Bring an idea, refine it in your engineering reality, and automatically create epics and tickets. Sign-up FREE.
Clicks’ Power Keyboard brings BlackBerry-style typing to any phone — with some compromises
aiste.ulozaite@gmail.com · unsubscribe
Clicks’ $99 Power Keyboard brings a customizable, slide-out physical keyboard to MagSafe and Qi2 smartphones, but its added heft can make larger phones awkward to use.
Launch your website faster than ever with Framer
AI Secret · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Framer is the pro website builder where AI agents work inside the product to make real, reviewable changes to your site. Agents can generate layouts, edit responsive breakpoints, refactor components, update CMS collections, and ship SEO-safe changes, with every edit visible on the canvas and under human control. Trusted by teams at companies like Miro and Perplexity.
MyClaw - Run Best-in-Class AI Agents
MyClaw Newsletter · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
MyClaw is a best-in-class AI agent platform that brings together OpenClaw, Hermes Agent, and commercial infrastructure to help businesses launch intelligent agents that understand goals, use tools, automate workflows, and deliver finished results.
Swap 30 min meetings for 3 second answers
uloza+hey@proton.me · *Sponsored* · unsubscribe
Unblocked augments source code with discussions from GitHub, Slack, Confluence, and more, so your team gets accurate answers about your codebase instantly.
Cloud vector databases were priced for humans. AI agents just broke the model.
Programming Digest · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Your agents query your vector database thousands of times per session. Cloud providers charge for every one. VectorAI DB runs on your own infrastructure with no per-query pricing and no egress fees. On-premises, at the edge, or air-gapped. Free to start.
Education
MIT’s Own AI Syllabus
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
13 courses walk through AI basics, machine learning, vision, and algorithms, and most of them cost nothing to take.
Health
Fit Senpai
TAAFT - There’s An AI For That · via aiste.ulozaite@gmail.com · unsubscribe
builds a workout plan around your level, matches it with meals and macro targets, then turns the week’s recipes into a grocery list.
Mayo Clinic faces lawsuit over MAYA assistant
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Mayo Clinic faces a lawsuit from a former AI compliance lead who says its MAYA assistant hid error-rate and privacy problems.
Insurers Push AI Into Care Decisions
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Health insurers are moving AI beyond post-care claims processing into earlier decisions about patient navigation, treatment recommendations and pre-approvals. Sun Life uses Medzown to steer complex cases toward clinical trials, while Aetna personalizes care pathways. As adoption expands, seven states have imposed guardrails requiring human clinicians to approve denials or delays in care rather than AI alone.
Coding
The End of Cursor
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Cursor, the AI coding IDE built by four MIT grads into $4B annual revenue and 50,000 enterprise customers, is being acquired for $60 billion The acquisition will result in Cursor’s name disappearing, with its unreleased agent shipped under a Grok label instead Cursor’s model-agnostic approach attracted users away from Copilot, but the acquisition demonstrates that app-layer companies without their own AI models cannot remain independent
GitHub added Copilot side chats and isolated worktrees
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
GitHub added Copilot side chats and isolated worktrees so developers can run parallel AI coding sessions more safely.
Ads
Time Is Stuffing Ads Into AI Memory
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Time magazine is serving hidden ads to AI crawlers in FAQ-shaped article versions visible only to machines like ClaudeBot These ads are designed to shape AI model outputs and long-term memory rather than reach human readers Publishers are pivoting to advertise to machines because human web traffic has declined and bot traffic now dominates their audience
Retail
Retailers Still Own the Wallet
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Retailers are racing to appear in AI chatbot recommendations as shoppers increasingly ask ChatGPT and Gemini what to buy Visitors sent by AI spend 41% more per visit, but retailers steer actual purchases back to their own sites to retain customer data and loyalty OpenAI shut down in-chat checkout, retreating to discovery; trust-based checkout still requires the decades of returns and loyalty programs only established retailers possess
Two-dollar Canadian Tire bags sell out after celebrity carries one
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Celebrity sighting of Canadian Tire bag sparked selling out Star from ‘Heated Rivalry’ carried the $2 bag Unexpected viral moment for basic retail product
Two-dollar Canadian Tire bags sell out after Heated Rivalry star is seen carrying one
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Canadian Tire bags became hot commodity after celebrity endorsement Inexpensive product ($2) sold out due to influencer effect Example of unexpected viral retail moment
Social Media
X is replacing revenue sharing with Original Content Rewards
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
X is replacing revenue sharing with Original Content Rewards to reduce incentives for AI spam and clickbait.
Tech
Cloudflare launched Kitesurf
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Cloudflare launched Kitesurf, an agent-first browser built on Workers V8 isolates for lighter automated browsing workloads.
Roll out AI-generated code progressively and adjust in real time
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
AI may be overwhelming your SDLC, but you can keep up. With LaunchDarkly CodeControl, you own what ships down to the last detail. Control exposure with targeting, segmentation, and progressive rollouts so you can make precise adjustments in real time. Automatically detect and remediate issues as they happen. Roll back or adjust changes instantly without manual intervention or 2 a.m. fire drills.
Apple Explores Shaking Up Its Smartwatch for a New Era
aiste.ulozaite@gmail.com · unsubscribe
The Apple Watch continues to attract new buyers and remains the top-selling smartwatch. However, the company’s industrial design team has been exploring a broad rethink of Apple’s smartwatches and is evaluating several different directions. Apple needs to adapt to the rise of screenless fitness bands and smart rings, which are cheaper, lighter, and more fun and convenient to use for many people. The company aims to inject new life into its Wearables, Home, and Accessories segment, which has struggled to generate meaningful growth in recent years.
Now we have a timeline of the OpenAI accidental attack against Hugging Face
aiste.ulozaite@gmail.com · unsubscribe
OpenAI recently gave a presentation about the Hugging Face incident. The talk is short and information dense and provides full details of what happened and how things played out inside OpenAI. This post presents the timeline of events and details how the hack occurred. A video of the presentation is available in the post.
China-Free Batteries Made From Salt Are Finally Here
aiste.ulozaite@gmail.com · unsubscribe
Peak Energy, a US-based startup, has raised tens of millions of dollars to build a factory for its sodium-ion battery packs. It already has more than $1.1 billion in announced deals with customers. Sodium batteries have less-intensive cooling needs than lithium-ion packs, and they are less likely to catch fire. Another US-based startup, Inlyte Energy, is working on another type of sodium-based battery that can’t catch fire at all.
Google’s Westinghouse Bet
aiste.ulozaite@gmail.com · unsubscribe
Google’s recent reshuffling of its AI operations has many thinking that the company has failed in the AI race. However, the company may just be refocusing its resources on its strengths. Google Cloud may be a much larger economic opportunity than pursuing its rivals in the frontier AI race. The company’s enormous distribution advantage means it doesn’t have to compete in the expensive and unprofitable frontier race.
Amazon cracks down on ‘CPU waste’ among engineers as agentic AI crunch intensifies
aiste.ulozaite@gmail.com · unsubscribe
AWS is cracking down on internal use of EC2 instances among its engineers. Engineers have been told to reduce CPU waste to ensure the company has enough CPU capacity to meet customer demand. The increased demand for CPU is due to the rise of AI agentic workloads. Other cloud companies are also seeing high demand for CPUs.
Interviewing Engineers in the AI Era: Lessons from a Year of Rebuilding
aiste.ulozaite@gmail.com · unsubscribe
Coinbase has rebuilt its engineering interview loop to test how candidates direct AI, evaluate its output, and apply judgment where models fall short.
Shopify’s blowout earnings show AI tools are paying off
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
While other tech giants struggle to justify their AI spending, Shopify is seeing major benefits. The company’s stock surged by 25% on Wednesday after it beat Q2 revenue-growth expectations (its fifth consecutive quarter of 30%+ growth), partly thanks to its AI tools — like one that helps merchants swiftly build their online storefronts. Its strong revenue from payment processing also strengthened its case to investors that it controls the plumbing of the e-commerce economy.
Myspace is coming back
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Social media platform Myspace announced revival plans Nostalgia driving interest in early 2000s social platforms Competes with modern social media landscape
Shopify’s blowout earnings
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Shopify’s stock surged 25% after beating Q2 revenue-growth expectations (its fifth consecutive quarter of 30%+ growth), partly thanks to AI tools that help merchants build storefronts and strong revenue from payment processing.
Myspace is coming back
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Myspace platform is being revived or relaunched Nostalgia-driven return of early 2000s social network Tests whether defunct platforms can regain relevance
Embattled hedge fund Situational Awareness invests $400M in chip startup Source Foundry
aiste.ulozaite@gmail.com · unsubscribe
The AI-focused hedge fund is still making some big bets.
Cloudflare Builds Browser for Agents
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Cloudflare launched Kitesurf, a cloud-hosted browser built specifically for AI agents, letting developers automate web navigation, forms, screenshots and HTML extraction without building their own browser stack. It runs on Cloudflare Workers and is designed to cut CPU, memory and token costs. Kitesurf is free during beta.
YouTube now requires creators to have twice as many watch hours to start earning money
aiste.ulozaite@gmail.com · unsubscribe
Creators who want to start earning on the platform will need at least 8,000 qualified watch hours over the past year or 20 million qualified Shorts views in the last 90 days.
Archer buys former rival Wisk Aero
aiste.ulozaite@gmail.com · unsubscribe
The two companies were once embroiled in a trade secret theft lawsuit. Now, Wisk is being absorbed into Archer.
Linux is seeing larger kernel update waves
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Linux is seeing larger kernel update waves as AI-assisted review helps developers push more small fixes into the pipeline.
Google’s classic Search button is gone in a new AI-first homepage
aiste.ulozaite@gmail.com · unsubscribe
Google is testing a version of its homepage where the search button is replaced with an AI Mode button. The new interface allows users to ask about files and use the Brainstorm function without having to log in. Logged-in users are also able to generate images. Ask about files allows users to attach a file before submitting a question, and Brainstorm helps users explore ideas.
Nvidia lines up $500 billion in financing as CEO Jensen Huang tells CNBC his chips are ‘investable asset’
aiste.ulozaite@gmail.com · unsubscribe
Nvidia has partnered with six large asset managers on a $500 billion financing push designed to treat compute infrastructure much like commercial real estate, toll roads, or other assets to borrow against. The effort will mobilize third-party capital for hyperscalers, frontier AI labs, and enterprises to build out data centers and acquire Nvidia hardware. It will help Nvidia’s customers secure financing without tapping their own balance sheets. Nvidia believes lenders can reliably underwrite compute as a revenue-generating asset with an extended life as its hardware is broadly adopted and transferable across customers.
Following Epic loss, Google has started hosting rival app stores in the Play Store
aiste.ulozaite@gmail.com · unsubscribe
Aptoide, a game-focused store that has been available as a sideloaded client since the early days of Android, is the first third-party app store to arrive in Google Play.
Apple is working on a way to authenticate that a photo came from an iPhone camera
aiste.ulozaite@gmail.com · unsubscribe
Apple Reference Image is designed to authenticate the source of photos using unique data tied to the iPhone camera hardware that captured them.
OpenAI reportedly completed a $7 billion employee tender offer
aiste.ulozaite@gmail.com · unsubscribe
San Francisco’s housing market is in trouble again.
Aptoide becomes the first rival app store to return to Google Play in the US
aiste.ulozaite@gmail.com · unsubscribe
Aptoide has brought its games store back to Google Play after more than a decade, as court-ordered changes open Android to competing app stores.
Spotify will label ‘AI Persona’ profiles and exclude their music from recommendations
aiste.ulozaite@gmail.com · unsubscribe
Spotify is introducing “AI Persona” labels for artist profiles that represent AI-generated identities and will exclude their music from editorial, algorithmic, and personalized recommendations by default.
How Cloudflare Is Making AI Pay for Content
ByteByteGo · via aiste.ulozaite@gmail.com · unsubscribe
Cloudflare’s reverse proxy position allows it to classify automated traffic and charge for requests from AI agents and crawlers The company evolved from blocking bot traffic to implementing Pay Per Crawl (flat per-request pricing) and Pay Per Use (pricing by outcome) Cloudflare resolves identity, permission, and payment within a single request using Web Bot Auth for cryptographic verification and the x402 protocol for payment
Optimizing Write-Intensive Database Performance (Sponsored)
ByteByteGo · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Free masterclass: Learn practical strategies for predictable low-latency writes at scale
Software engineering at a proprietary trading company: Optiver
The Pragmatic Engineer · via aiste.ulozaite@gmail.com · unsubscribe
A shift from a focus on latency to building better AI models, owning the full stack from applications to building custom hardware, very different incentives to most tech companies in play, and more
Agents on Rails
uloza+hey@proton.me · unsubscribe
New Rails AI initiative for agents Includes announcement and first benchmark report Read the announcement and benchmark report for details
Add support for the HTTP QUERY method
uloza+hey@proton.me · unsubscribe
QUERY is a safe and idempotent HTTP method that conveys the query in the request content, making it suitable for queries too large or structured for a URL query string
Show bin/console startup banner based on IRB
uloza+hey@proton.me · unsubscribe
Show a Rails-flavored startup banner (a small logo, Rails/Ruby version, a rotating tip about console helpers like reload!, and Rails.root) when starting bin/rails console. Set RAILS_TIPS=false to hide the rotating tip.
Allow ffmpeg and ffprobe input arguments to be configured
uloza+hey@proton.me · unsubscribe
Two new configuration parameters are introduced. config.active_storage.video_preview_input_arguments is passed to ffmpeg before -i, and config.active_storage.ffprobe_arguments is passed to ffprobe before the file path. Both default to empty string.
Allow config.active_storage.variant_processor to be set to a class
uloza+hey@proton.me · unsubscribe
The class must implement the interface defined by ActiveStorage::Transformers::Transformer. Active Storage then uses it for variant processing
The schema readers answer for many tables at once
uloza+hey@proton.me · unsubscribe
indexes / primary_keys / foreign_keys / check_constraints / exclusion_constraints / unique_constraints now accept a list of tables and answer for all of them at once
Fix LengthValidator crash with proc minimum and nil value
uloza+hey@proton.me · unsubscribe
The proc was resolved only when the value was present, so for a nil value it leaked unresolved into the error message and was invoked with the message options hash instead of the record. It is now resolved before the error is built, producing the expected “is too short” message.
Use bind parameters for array-form arguments in find_by_sql / count_by_sql
uloza+hey@proton.me · unsubscribe
The API doc claims parity with where, but array-form values were eagerly interpolated via sanitize_sql.
Deprecate create alias of insert in connection adapters
uloza+hey@proton.me · unsubscribe
insert / update / delete correspond to their SQL DML verbs, but create reads like DDL (create_table / create_database) rather than the INSERT it actually performs.
Split keyword arguments off #args on Rails.application.middleware entries
uloza+hey@proton.me · unsubscribe
Middleware entries now expose keyword arguments through a new #kwargs accessor. #args previously bundled kwargs as a trailing hash inside the positional array (via Hash.ruby2_keywords_hash). It now returns positional arguments only.
Allow to configure ERB options through ActionView::Base
uloza+hey@proton.me · unsubscribe
The ActionView::Template::Handlers::ERB class is now private API. Applications that used to configure ERB options such as escape_ignore_list now need to do this on the ActionView::Base class or on the railtie config.action_view configuration.
Append TRADITIONAL instead of STRICT_ALL_TABLES to MySQL’s sql_mode by default
uloza+hey@proton.me · unsubscribe
On environments whose global sql_mode is empty — most notably Amazon RDS and Aurora MySQL default parameter groups — appending only STRICT_ALL_TABLES leaves out NO_ZERO_IN_DATE, NO_ZERO_DATE, and ERROR_FOR_DIVISION_BY_ZERO, which MySQL 5.7+ has otherwise made part of its own default. Appending TRADITIONAL closes the gap.
Allow ActiveSupport::ProxyLogger to ignore messages by pattern
uloza+hey@proton.me · unsubscribe
Useful to silence noisy logs from gems that your application may not care about, without needing to change the log level and losing other useful logs.
Add Pacific Time (Canada) and Alberta Time zone mappings
uloza+hey@proton.me · unsubscribe
Adds two friendly zone mappings: Pacific Time (Canada) → America/Vancouver, Alberta → America/Edmonton
Change the shape of ActiveRecord::Migration::CommandRecorder#commands
uloza+hey@proton.me · unsubscribe
Each recorded migration command is now stored as [cmd, args, kwargs, block] (4-element) instead of [cmd, args, block] (3-element) with kwargs bundled into a trailing hash inside args. Code that inspects recorder.commands directly needs to adapt to the new tuple shape.
Fix normalizes to run before the underlying type validates an assigned value
uloza+hey@proton.me · unsubscribe
When normalizes was combined with another type that rejects invalid input (such as an Active Record enum), the underlying type’s assert_valid_value ran against the raw, un-normalized value and raised before normalization had a chance to run. The normalization is now applied first, so a value like “ Pending “ is normalized to “pending” and accepted by the enum.
Negative-interest tech debt
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Tech debt is usually prioritized by how fast it compounds, but Jesse points out that AI may be changing that calculation. If AI keeps improving, fixing messy code in the future could be easier than fixing it today, even after more code has piled on top. That makes the effective interest rate on tech debt potentially negative, which would mean shipping fast now beats cleaning up later. Companies are already betting on this, though nobody yet knows if the bet will pay off.
Nvidia’s Risky Business
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Why compute might get 10x more expensive in coming years
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Improving heuristics
Programming Digest · via aiste.ulozaite@gmail.com · unsubscribe
Placing “landmark” nodes on a map can make A* run faster without changing the algorithm itself. Before each search, Amit suggests running Dijkstra’s from each landmark and storing the costs. Then use the triangle inequality to get a tighter lower bound on path length, replacing the standard distance heuristic. Placing more landmarks near map edges or common paths means fewer nodes are explored, though the best placement depends on your specific game and map layout.
A deep (and fuzzy) dive into search
Programming Digest · via aiste.ulozaite@gmail.com · unsubscribe
Fuzzy search helps search engines find results even when words contain typos or extra characters. Andrey explains two approaches used by SereneDB: Levenshtein automata and n-gram similarity. Levenshtein is useful for precise typo matching, while n-grams work better with longer text and partial matches. SereneDB combines these techniques to make fuzzy search practical and flexible.
A shell exclamation mark is not for yelling
Programming Digest · via aiste.ulozaite@gmail.com · unsubscribe
Filip introduces shell event designators, a powerful but often overlooked way to reuse and modify previous commands. He explains useful shortcuts such as !!, !$, and !ssh, along with word and path modifiers. Filip also covers the POSIX fc command, safer ways to use these features, and when alternatives like Ctrl-R may be better.
Science
The first self-driving vehicle on Mars has proven to be a smashing success
aiste.ulozaite@gmail.com · unsubscribe
NASA’s Perseverance Mars rover will set a record for the most distance driven by any vehicle on another world sometime next week. The rover’s auto navigation system has enabled about 90% of the distance driven by the vehicle to be autonomous. Not having to wait for navigation commands from drivers on Earth has allowed the rover to maximize the amount of scientific return. The vehicle will likely be able to keep driving itself across Mars for many years to come.
Programming
Agentic Code Quality
aiste.ulozaite@gmail.com · unsubscribe
Software quality now depends on the constraints set around the agents that wrote it. Constraints define what a system is allowed to do. Setting and maintaining these constraints allows developers to build loops that reliably deliver high-quality production software. Agents can propose anything - constraints decide whether a proposal is safe enough, correct, scoped, and useful.
Changing Devtools Is Cheap. Owning Them Isn’t
aiste.ulozaite@gmail.com · unsubscribe
The future of software will be powered by high-quality, well-documented building blocks that users can build specialized artifacts on top of. Agents are good at composing high-quality components. Maintainers can provide these components alongside a focus application, and users can choose how to modify their experiences themselves. This allows software to be personalized while still being developed by maintainers that care.
Why you might want to build your WebApp in Canvas instead of HTML
aiste.ulozaite@gmail.com · unsubscribe
Canvas is a low-level rendering tool that gives users more control than HTML.
How I Rebuilt My Development Workstation for Agentic Work
aiste.ulozaite@gmail.com · unsubscribe
This post describes a workstation that is terminal-native, agentic-first, and built around composable tools.
What’s the best programming language for coding agents?
aiste.ulozaite@gmail.com · unsubscribe
Anyone who wants to make a strong claim about which language or classes of languages are particularly good with LLMs would need to run quite a few different evaluations. Failures with languages are generally something idiosyncratic, where it’s not always obvious how much the issue generalizes across tasks or setups. There’s no way to look at the score on one or even several evaluations and draw a conclusion about programming in general. At this point, it would be a mistake to draw a strong conclusion about any particular language.
Markets
The AI trade is blowing up — and July cleared out risky bets
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Chip stocks collapsed 26% in a month during summer AI trade concerns Situational Awareness hedge fund blew up, losing $35 billion 86% of U.S. companies beat earnings expectations, highest share since 2021, reigniting bull market
Surprise! Situational Awareness was the bottom
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
AI-focused hedge fund Situational Awareness collapsed in July, losing $35 billion Strong corporate earnings (86% of U.S. companies beat expectations) helped the market recover Chip stocks rebounded +17% from recent lows as investors refocused on profits rather than AI trade risks Major indexes reached fresh all-time highs following earnings surprises
Consumer
Gen Z’s ‘little treat economy’: $8 pastries instead of homes
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Gen Z doesn’t have much in the way of savings, but it does love to spend on little treats. Zoomers burn cash on beauty products, snacks, and jewelry. Economists are calling it the ‘little treat economy,’ the result of joyful splurging on frivolous purchases to fill the void left by unaffordable big-ticket items, like homes. Gen Z’s average spending on little treats grew by more than 6% YoY in June, compared to 3% for all generations.
Media
Disney signs deal to let TikTokers incorporate its content into posts
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Disney content can now be incorporated into TikTok creator posts Part of Disney’s broader content distribution strategy Expected to drive new creative formats on the platform
Auto
BMW puts Spider-Man movie ads on car dashboard screens
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
In-vehicle advertising for Spider-Man movie appears on BMW dashboards Example of automotive tie-in marketing with entertainment Targets drivers with ads during their commute
Biotech
Biotech startup creates gene-edited beagles for dog-allergic people
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Gene-edited beagles designed to be hypoallergenic Addresses market demand for dog-loving people with allergies Example of biotech application in consumer products
Biotech startup creates gene-edited beagles for people allergic to dogs
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Biotech startup uses gene-editing to create hypoallergenic beagles Addresses pet allergy concerns for dog lovers Innovation combining biotechnology with consumer pets
Business
Crocs keeps Malta office to dodge taxes
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
New York Times investigation uncovers Crocs’ Malta tax strategy Offshore office used for tax avoidance purposes Part of broader corporate tax minimization practices
NYT investigation finds that Crocs keeps an office in Malta to dodge taxes
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Crocs uses Malta office for tax avoidance strategies New York Times investigation uncovered corporate tax practices Highlights broader issue of corporate tax optimization
OpenAI completed a $7 billion employee tender offer
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
OpenAI reportedly completed a $7 billion employee tender offer at an $852 billion valuation.
OpenAI acquired NextSlide
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
OpenAI acquired NextSlide, adding prompt-to-presentation tooling to its broader ChatGPT productivity push.
See How a Tesla-SpaceX Merger Gives Musk a Shortcut to His $1 Trillion Payday
aiste.ulozaite@gmail.com · unsubscribe
The conditions for Elon Musk to receive his $1 trillion pay package at Tesla could easily be achieved if the company were acquired. There is a clause in the agreement that declares half the targets as good as accomplished if Tesla is acquired or otherwise taken over. This leaves only market-value targets to determine how many shares Musk gets. Musk has considerable say over any offer SpaceX makes for Tesla. Even a stratospheric bid would keep Musk in firm control of the combined company.
Compute Is Revenue. Revenue Is Collateral
aiste.ulozaite@gmail.com · unsubscribe
Nvidia is attempting to turn a handful of bespoke transactions into a financial production line.
Food
U.S. salad chain ‘Salad and Go’ closes all locations
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Fast-casual salad chain shutters all U.S. locations Health-focused restaurant concept failed to sustain Part of broader challenges in quick-service salad market
Fashion
Air-conditioned clothing debuts at Paris Fashion Week
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
New cooling technology integrated into garments shown in Paris Addresses climate change and wearable technology trends High fashion embraces functional cooling features
Crypto
How I Lost $1.6 Million in Bitcoin in 8 Minutes
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
On July 31, Toronto-based entrepreneur and personal trainer Jonathan Goodman, 40, discovered that he’d lost $1.6 million in a major Bitcoin hack. The thieves targeted offline-password devices made by the Canadian company Coinkite.
Consumer Trends
Gen Z’s $8 pastry habit
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
Gen Z doesn’t have much savings but loves spending on little treats. The trend shows Gen Z average spending on little treats grew by more than 6% YoY in June, compared to 3% for all generations, driven by the ‘little treat economy’ of joyful splurging on frivolous purchases to fill the void left by unaffordable big-ticket items.
Marketing
BMW is putting ads for the new Spider-Man movie on dashboard screens
TLDR by Wealthsimple · via aiste.ulozaite@gmail.com · unsubscribe
BMW incorporating Spider-Man movie ads on vehicle dashboard screens Novel in-vehicle advertising approach for film promotion Tie-in between automotive brand and entertainment content
Security
A data breach at shipping giant Ceva Logistics is rippling across banks, retailers, Steam gamers, and beyond
aiste.ulozaite@gmail.com · unsubscribe
Companies that rely on Ceva Logistics for shipping their physical goods to customers say their personal data was taken during a recent cyberattack.
Signed up for Klaviyo? Dozens of advertisers may have seen your password
aiste.ulozaite@gmail.com · unsubscribe
A bug in the tech giant’s website mistakenly shared users’ sign-up information, including personal data and their password, to third-party companies.
Everything I Learned Shipping Device Bound Session Credentials
aiste.ulozaite@gmail.com · unsubscribe
Device Bound Session Credentials bind sessions to a private key a browser generates in hardware and cannot export.
A researcher bought noreply.net. Companies started sending him secrets
aiste.ulozaite@gmail.com · unsubscribe
Cory Solovewicz created an accidental honeypot when he purchased noreply[.]us and noreply[.]net in 2020 and 2024, respectively.
North Korean remote IT staffer worked for US government agency, says FBI
aiste.ulozaite@gmail.com · unsubscribe
The investigation shows that North Koreans are able to infiltrate government agencies, as well as private organizations and crypto exchanges.
The Danger of Autonomous AI in Cybersecurity
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Investigating three real-world incidents in our cybersecurity evaluations
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
How Figma Stays Ahead of Vulnerabilities With Agents
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Email encryption
Programming Digest · via aiste.ulozaite@gmail.com · unsubscribe
Email was invented in many places at once, and its security was never a priority. Two main encryption standards emerged: PGP, which is open and decentralized but notoriously hard to use, and S/MIME, which works well inside organizations with shared infrastructure but struggles across them. In practice, most email today is encrypted in transit between servers, which solves many real threats even without full end-to-end encryption. The core problem is that strong encryption requires key distribution, and no universal solution for that has ever taken hold.
Mobility
TechCrunch Mobility: Zoox prepares for launch and Uber’s AV empire
aiste.ulozaite@gmail.com · unsubscribe
Welcome back to TechCrunch Mobility, your hub for the future of transportation and now, more than ever, the role AI is playing in it.
Commerce
Retailers Want AI Traffic, Not Checkout
MyClaw Newsletter · via aiste.ulozaite@gmail.com · unsubscribe
Retailers are embracing ChatGPT and Gemini as new shopping discovery channels, with AI-referred visitors already converting better than traditional traffic. But companies such as Ulta, Etsy and The Knot still want customers to complete purchases on their own sites, where they retain browsing, purchase and loyalty data. The goal is simple: let AI recommend, but keep the customer relationship.
Sponsored
Invest in High-Potential Startups Like These
aiste.ulozaite@gmail.com · *Sponsored* · unsubscribe
Alumni Ventures is giving readers early access to high-potential startup opportunities across today’s most active sectors, co-invested alongside name-brand VC firms like Andreessen Horowitz (a16z), Bessemer, & Y Combinator.
Politics
Google co-founder Sergey Brin has now spent $100 million to fight the billionaire tax
aiste.ulozaite@gmail.com · unsubscribe
California’s Prop 40 would impose a one-time 5% tax on the net worth of the state’s billionaires.
Defense
Sila lands $1.4B Pentagon loan as militaries demand more batteries
aiste.ulozaite@gmail.com · unsubscribe
Battery materials startup Sila will use a $1.4 billion loan from the U.S. Department of Defense to scale production at its factory in Washington state.
Compute
Nvidia’s $500B Debt Flywheel
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Nvidia partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to raise $500B for AI data centers GPUs repositioned as ‘AI factories’—investable assets with revenue generation and resale value like power plants Nvidia backstops up to 25% of residual project value, echoing dot-com vendor financing models that collapsed when real demand fell short
Microsoft is reportedly preparing to unveil Maia 300
AI Secret · via aiste.ulozaite@gmail.com · unsubscribe
Microsoft is reportedly preparing to unveil Maia 300 this fall, pushing deeper into custom AI chips to reduce Nvidia dependence.
Robotics
Dyna-2 Proves Scaling Laws for Robotics: 1 Million Hours of Human Video Unlocks Zero-Shot Dexterity
aiste.ulozaite@gmail.com · unsubscribe
Dyna-2 is a world-action model (WAM) pre-trained on over one million hours of human video data. Its existence proves that scaling human video data predictably improves zero-shot performance on unseen robot hardware. Dyna-2 achieved an 87% pass rate in real-world zero-shot deployments at customer sites, vastly outperforming the 46% pass rate of its VLA predecessor, Dyna-1. The model uses a new one-step video generation distillation pipeline that drops inference latency by two orders of magnitude for downstream planning.
CodeReview
Review The Actual Change, Not The File List
Pointer · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
AI writes more code than ever. Reviewing it shouldn’t mean scrolling forty files in alphabetical order. CodeRabbit Review reorganizes any pull request from a flat file list into a structured, layer-by-layer walkthrough - the logical reading order of the change, not the order your platform happens to sort it. Every range gets its own plain-language summary, with sequence diagrams, state machines, and ERDs generated inline wherever a visual earns its place. (1) Cohorts group related files and chunks so you review one idea at a time. (2) Layers order them so foundational changes - data shapes, contracts - come before the code that depends on them. (3) Code Peek lets you click any variable, function, class or type to see its definition and usages without leaving the tab. (4) Semantic Diff cuts past formatting noise to show what actually changed. From the team that pioneered AI code reviews. 2M reviews every week. 6M repos. 15K customers. Free during early access.
Leadership
Relentless Forward Progress
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
The hardest stretches of running a team are not sprints. They are ultras: the long downturn, the slow turnaround, the market that stays bad for six quarters straight. None is solved by a single heroic burst. They are survived, or not, by how you manage the long middle, when the adrenaline is gone and the only skill left is suffering productively while metering what you have.
The same side of the table
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Good managers always stand behind their team, even when things go wrong. Throwing someone under the bus destroys trust and creates a culture of fear that hurts future performance. Instead, step in gracefully, redirect the conversation, or take ownership of the situation. The best people want to work with leaders who have their back.
The goal is (still) value
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
AI adoption and higher output do not automatically mean more value. Francisco says that engineering teams should measure what really matters: customer outcomes, delivery speed, and the investment needed to achieve them. Instead of focusing on AI usage, prompts, or token spending, leaders should connect metrics to cycle time and cost. The goal is simple: deliver useful results faster and with less investment.
Middle management roles are a trap
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe
Growth
The Antithesis Principle
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
(1) If you want to help people learn something, look into making it engaging / entertaining. This first one is the easier, more obvious conclusion. (2) It is actually not in your best interest to be the kind of person who can only learn from entertaining content. Nor to be the kind of person that needs any sort of entertainment as a prerequisite for learning. The vast majority of smart people will never consider this second conclusion.
Management
Middle Management Roles Are Also A Trap
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
If you’re willing to embrace this fact that a little bit of time in middle management roles is important preparation to become an executive, but spending a great deal of time in middle management only prepares you for further middle management roles–and makes you less effective as a potential executive–then the AI-driven shift in managerial fads might be a threat to your current role, but it’s likely to improve your chances to succeed as an executive.
Onboarding
A Practical Guide To Rapid Onboarding
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
After nearly two decades at Amazon, operating as a Principal Engineer, often parachuting into complex, high stakes projects, I’ve developed a system to cut through this chaos. In this article, I’ll share with you that exact framework to help you get up to speed unbelievably fast.
Auth
Agents Can Now Sign Up For Your App
Pointer · *Sponsored* · via aiste.ulozaite@gmail.com · unsubscribe
Agents hit your signup flow and bounce off a login built for humans. Every one that gives up is a signup you never see. WorkOS Agent Registration turns that traffic into signups. Enroll via the dashboard and AuthKit publishes an auth.md file agents read to register for scoped, short-lived credentials you control.
CareerGrowth
Growing As An Engineer In A World Of AI
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
There’s a lot of buzz that AI means companies only want seniors now. That there’s no room left for early career engineers. I don’t think that’s true. The thing actually worth worrying about isn’t whether you’ll get hired, it’s whether you keep learning in a world of AI, now that it removes the friction where that learning used to happen. And that, thankfully, is a choice you get to make.
Databases
Beyond Happy Path Engineering: Databases
Pointer · via aiste.ulozaite@gmail.com · unsubscribe
This series is about making your programs more resilient. Each post looks at one part of the system we take for granted: networks, time, databases, storage, users, dependencies, deployments, queues, and concurrency. The question is always the same: what assumptions are we making here, and what should we do when they stop being true?
Peopleware
Being a Staff Engineer at Stripe in 2026
uloza+hey@proton.me · unsubscribe
A lot of it is generic and does not provide practical insight to use, still, one thing that is important to get comfortable (and learn to find balance) with is this: “There is a genuine pressure to get things done quickly, because there are always 10 more things in the pipeline waiting to get done.”
Using Agents Is Literally Engineering Management
uloza+hey@proton.me · unsubscribe
Wonderful post by Jack Danger with a great take that truly captures the early days in managing humans and AI: “A first-time manager cuts hard toward either micromanaging or abdication… That’s the IC-and-AI spectrum right now. Either you hand-hold every line of output, or you YOLO it to production and hope. […] let yourself accept that all of the models and all of the people are imperfect. Then commit to building the skills.”
A Fresh Definition of the Product Role
uloza+hey@proton.me · unsubscribe
If the cost of building is lower, the value of knowing what to build is much higher. The problem is that many product managers think they’re great because they can articulate what to build. Great PMs can deeply articulate why to build: “the hard part is knowing that it should exist, and knowing how it should exist.”
Industry
Why Is Everyone In Tech So Sad?
Leadership in Tech · via aiste.ulozaite@gmail.com · unsubscribe