Your AI app wants your fingerprints
A ‘free’ app asks for your fingerprints. Meanwhile, the web forgot how to talk to AI. Trust is the real bottleneck. We’re building, testing, and keeping our data close. Esto te toca — la gente needs to know what’s asking for what.
Your AI strategy has a trust problem
The headline says it plainly: your AI strategy has a trust problem. That’s not a branding issue. It’s an operational one. People don’t use AI tools when they don’t trust them. And trust isn’t built by adding more features or lowering prices. It’s built by showing how the tool works, who benefits, and what goes wrong when it doesn’t.
The author, writing on Substack, frames this as a product strategy issue. The advice is practical. Don’t hide behind jargon. Don’t promise magic. Show the inputs, show the outputs, show the limits. Let people test it themselves. Let them see the seams. Trust comes from transparency, not from hype.
This hits hard for small business owners, freelancers, and community organizers who are already being pushed to adopt AI tools by consultants or vendors. If the tool doesn’t explain itself, people won’t use it. Period. And that’s a problem for anyone trying to build a real product, not a demo.
The post doesn’t give a step-by-step guide. It’s more of a wake-up call. But the point is clear: trust is the bottleneck. Not cost. Not speed. Trust. And trust is earned by being honest about what the tool can and can’t do.
Why this matters for us: When we try to use AI to help our businesses or our communities, we need tools we can trust — not flashy demos that vanish when the novelty wears off.
Why one-shot planning always fails teams
TLDR Product's latest issue points to a simple, boring truth: teams that plan once and forget it fail. The article links to Atomic Object's research on in-person backlog management, where the key finding is that teams doing face-to-face planning sessions deliver more reliably than those relying on solo planning or async tools alone.
The setup is familiar. You schedule a two-hour sprint planning. Someone brings a spreadsheet. Someone else brings a whiteboard. You get through the list. Then life happens. The plan is out of date by Tuesday.
Atomic Object's data shows that in-person, collaborative backlog grooming — where the whole team sees the work, argues about priorities, and commits together — produces better estimates and fewer surprises. Not because the people are smarter. Because they're talking to each other.
The one-shot model looks efficient. You do the work once, you move on. But planning is not a ceremony. It's a conversation. And conversations break when you stop having them.
This isn't about tools. It's about whether your team actually sits in the same room — or the same Zoom — and decides together what matters next. The spreadsheet doesn't decide. The spreadsheet doesn't notice that the client changed their mind. The spreadsheet doesn't know your cousin's side hustle needs that feature first.
Why this matters for us: most of our people are stuck doing solo planning in a Slack thread while the whole team is scattered across time zones — so the plan dies before it's even built, and we pay for it in overtime and broken trust.
Fine-tuning: When AI Gets the Tacos Right
You know that big AI chatbot that can write essays, code, and answer questions? It's a generalist. Good at everything, great at nothing. Fine-tuning is when you take that generalist and train it on your own stuff — your family recipes, your shop's inventory, your community's rules.
Think about your abuela's tamale recipe. She doesn't read a cookbook. She learned from her mother, from the neighborhood, from trial and error at the family gathering. The recipe got fine-tuned through generations. The AI does the same thing, just faster.
Here's how it works in plain terms:
The base model is like a new hire at the bodega. He knows how to ring up a sale, but he doesn't know your customers' names, what they usually buy, or which shelf the special is on. Fine-tuning is showing him your register for a week. Suddenly he's useful.
You don't need to be a coder for this. Some platforms let you upload a PDF of your business guidelines, a folder of your team's notes, or a collection of your customer service emails. The AI reads them, adjusts its responses, and starts talking like your voice instead of a generic corporate one.
The catch? If your training data is messy, the AI gets messy. Garbage in, garbage out. Your auntie's Facebook posts about conspiracy theories won't make a good training set for a business bot.
Why this matters for us: If your business, your church, your neighborhood org, or your familia's group chat has a lot of shared knowledge, fine-tuning lets you build an AI that actually speaks your language — not the Silicon Valley version. It keeps the culture in the machine.
Tip: Before you fine-tune, ask yourself — what's the one question your community asks you every single day? Start there. One good answer, cleaned up and repeated, beats a thousand random documents.
The web is not the AI. The web is for people.
— paul.kinlan.me
#the-web-isn-t-dying-it-s-just-forgotten-how-to-talk-to-ai-1ed6ffSony says next PlayStation won't just live in the living room
Sony is talking big about the next PlayStation. In a recent investor Q&A, the company hinted its next console will offer a way to play games outside the living room.
The question was blunt: how do you bring back players who jumped to gaming PCs during the pandemic?
Sony’s answer? They’re moving beyond the couch. The exact details are sparse, but the direction is clear — the next PlayStation will be a portable experience.
This isn’t a total surprise. Sony already lets you stream PlayStation games on mobile devices. The rumored handheld, codenamed “Project Q,” is a recent addition to that lineup.
TradeWork: One app for the truck, not the office
You're out on a job site, phone in hand, trying to remember who owes what, what the crew is doing, and whether the estimate got sent. Your notes app is full of estimates. Your texts are full of crew updates. Your invoices are PDFs you emailed last Tuesday. And the check still hasn't cleared.
The big CRMs are too heavy. The trade apps are built for the office, not the truck. So you're juggling three apps and a paper clipboard, and the foreman is yelling because nobody read the job specs.
TradeWork fixes that. It's a mobile-first platform for trades crews who work out of a truck. Jobs, crews, invoices, and payments — all in one place. And it's bilingual, so the office, the foreman, and the helper can all read the same job. No more guessing who's on the job. No more chasing payments. No more losing estimates in a notes app.
This is for painters, plumbers, electricians, landscapers, and anyone who runs a trades business. Get it working on the job, not in the office.
https://tradework.work
Google Home Speaker Gets a Star Fox Upgrade
The Verge is wrapping up its Installer newsletter series, and the latest edition lands on a surprisingly specific tech pick. David Pierce calls it "the new best smart speaker" — a nod to Google Home's fresh Star Fox skin.
It's a small thing, sure. But it's the kind of update that shows up on a Tuesday night, right when you're loading the dishwasher and the kids are still on the couch. You tap it. It plays. You don't have to think about it.
That's the whole point of the thing. It's not a flex. It's not a "disruptor." It just does its job while you're busy being a person. A person with a kid who keeps asking for water at 11 p.m.
Why the 'free' app wants your fingerprints
A popular facial recognition app is asking users to scan their faces to unlock features — and it's not just a selfie. The app collects biometric data, including facial geometry, to build a unique identifier for each user.
The catch? That data isn't just stored locally. It's sent to a cloud server, where it's linked to your account, your location history, and your purchase patterns. The app's privacy policy is buried in a 40-page document most people never read.
What makes this worse is the opt-out. You can't use the app without granting biometric access. There's no alternative login. No way to use it "just for fun." You scan or you don't use it.
The company behind it, a startup that raised $12 million last year, says the data is "anonymized." But anonymized facial geometry can still be matched to a person using other databases. Several independent researchers have already shown this is possible.
Regulators are starting to ask questions. The FTC has opened a preliminary inquiry into whether the company is complying with existing biometric privacy laws. But there's no federal law that specifically bans this kind of data collection yet.
The app's founder told us the data is "useful for personalization." But personalization doesn't require a permanent biometric fingerprint stored on a server you don't control.
Why this matters for us: Our faces are already being scanned by police drones and store cameras. Now a startup wants to turn our biometrics into a product — and we don't even get to say no.
A side project built with AI: dbt + ClickHouse
A LinkedIn post shared by TLDR Data points to an open-source project that uses AI to help structure dbt models on top of ClickHouse. The post is by a developer named Oleg Agapov, who says he built the project to share how he structures data pipelines and dbt models with AI assistance.
The post links to a public repo and a blog post on ClickHouse’s site about open-source. The post itself is a personal share — a dev talking about how he puts together a project that combines dbt (the data transformation tool) with ClickHouse (the database). He says AI helped him organize the structure.
Home batteries: how they work, what they cost
A lot of folks are adding batteries to their homes. Not just for backup power. But to save money on electricity bills and ride out blackouts.
A home battery is a big lithium-ion pack, like the ones in EVs, but built for the house. It stores power when it’s cheap. It discharges when rates spike. Some hook into solar panels too. Others just ride the grid.
The cost? Usually between $10,000 and $20,000, installed. That’s a lot of pesos. A lot of quincenas. But prices have been dropping. And with some states offering rebates, the math can swing.
Installation isn’t plug-and-play. You need a licensed electrician. Sometimes a panel upgrade. Sometimes a new inverter. A lot of folks skip the electrician because they know a cousin who does “electrical work.” Don’t. That’s how you start a fire.
What to look for:
- Capacity (kWh). How much power it holds.
- Power output (kW). How fast it can give power back.
- Warranty. Some last 10 years. Some less.
- Compatibility. Does it work with your solar setup? Your inverter? Your utility?
Some batteries talk to the grid. They sell power back when rates are high. That can pay for part of the battery over time. But it depends on your utility. Check first.
The upside? You’re not at the mercy of the grid. When the line goes down, your lights stay on. Your fridge stays cold. The baby’s formula doesn’t spoil.
The downside? It’s expensive. It takes space. It needs maintenance. And if you buy the wrong one, you’re stuck with a brick.
Why this matters for us: Brown families in the South and Southwest are on the front lines of grid failures and rate spikes — a battery is one of the few ways to take control of your power, your bills, and your home’s safety without waiting on someone else’s grid.
Founders Fund backs a robot that kills fish humanely
Founders Fund is backing Shinkei, a startup building a refrigerator-sized robot called Poseidon. The machine is designed to kill fish quickly and humanely. It is not a consumer product. It is a tool for the fishing industry.
The idea is simple: reduce suffering at the moment of death. The fish industry kills billions of animals every year. Most methods are fast, but not always clean. Shinkei wants to fix that with a machine.
The robot is compact. It fits in a refrigerated space. It works alongside the existing workflow. It does not replace the worker. It handles the final step.
Founders Fund is known for big bets.
The marketing critical thinking gap: how to spot risky AI use cases fast
TLDR Marketing ran a Maven chat with Dan from TLDR Marketing about AI governance for CMOs. The gist is simple: marketing teams are moving fast on AI, and not everyone is asking the right questions yet.
The article points out a gap. Marketing people are eager to use AI tools. They are not always pausing to think through the risks. That is where the critical thinking gap comes in. It is not about being anti-AI. It is about being intentional.
The advice is practical. Before you launch a campaign with AI, ask what could go wrong. Who could get hurt? What data is involved? Is the output accurate? Can you explain it to your boss? Can you explain it to la gente who might see it? If you cannot answer those, do not ship it yet.
The article also talks about governance. CMOs need guardrails. Not heavy rules that slow everything down. Light guardrails that catch the risky stuff before it goes public. Things like approval steps, data checks, and clear lines on what AI can and cannot do in your brand voice.
This is not a lecture. It is a reminder that speed without thinking gets messy. And messy campaigns get called out. On Twitter. In the news. By customers who feel misled. That is the cost.
The tone is calm. No hype. No fear-mongering. Just a straight talk about how to keep your marketing sharp while still moving fast. The writer is not a lawyer. He is a marketer who has seen what happens when teams skip the thinking step.
Why this matters for us: when marketing teams skip the thinking step, la gente gets misled, and the people who pay the price are the ones who trust the brand the most.