Our team of ~8 devs is switching from entirely Claude Code to a mix of different agents with a ChatGPT subscription providing the inference. No idea if this is better or worse, but I *love* what this says about Anthropic/OpenAI's future. Neither one will ever be able to charge a premium.

It's neat the Bluesky has a way for accounts to label themselves as bots, but the natural next step is to let me block bots from replying to my posts. Why would I ever want a reply from a bot? If my post is meant as a question for AI, I would have just asked it to an AI directly.

I love the whiplash of SaaS companies going from "pay us double to use AI" to "please please please use this free AI feature".

I spent all yesterday alone in the office upgrading our wifi and rearranging the desks. It was a lot of fun and reminded me of the old days. But it also reminded me that in the old days, I spent way too much time on crap like that. I was so excited to "build a business" that I let it distract me.

As tech people, when designing features, we tend to get overly clever. But then when you talk to normal people using the software, the things they actually care about tend to be way more basic than you'd ever expect. Keep it simple.

I really with Bluesky had a "location" field on user profiles the way Twitter does. So many posts are directly or indirectly tied to a person's region (weather, traveling, taxes, regulation, etc.) and it's nice to be able to quickly see where they are so I can contextualize the post.

This page is a fun example of Google sucking: developers.google.com/identity/bra... The top section recommends using the first image as the logo. The next section shows the second image, a "G" with a gradient. Below that, it says that the first image is using an outdated logo which isn't allowed!

We set up Ring cameras at the office after there was a break-in, and wow is the admin panel terrible. So many things are straight-up broken. Redirects pointing to redirects. I can't even subscribe to a plan because it redirects me back to my dashboard every time I hit the billing page.

Fuck making money, this is what it's about: "As a long time user of LACRM I couldn't be more impressed with how you do business. What other company in America has customer support that actually answers the phone with a real live person. You are always improving an already great product"

Airtable is getting acquired by PE and presumably many employees will get laid off. Sad, but this is inevitable when you raise a billion dollars. Best case is employees get to cash out before the company enshitifies, but enshitification is coming one way or another. That's the deal when you raise $

Anil Dash@anildash.com · 4w ago

Terrible news, especially for their workers. Just a deeply disappointing ending for such a promising company. I hate this.

I don't understand people who *must* have dark mode for every minor site/app. I prefer light mode, but when a site defaults to dark mode it's totally fine. Not even important enough to switch if that's an option. Such a weird thing for so many people to care about.

I'm at a conference in Minnesota and I'm learning that everyone in Minnesota thinks St Louis is "the south". St Louis thinks that both places are the Midwest.

We just launched HIPAA compliance! Super tedious to implement, and maybe I'll regret it, but I'm cautiously optimistic. Our #1 industry is health insurance (about ~15% of customers) and now that we're the cheapest HIPAA-compliant CRM in the world (that I can find), I'm hoping we can do even better.

We're getting hit with a signup spam bot that caused us to send a few thousand emails to random addresses before we caught it. Very annoying, but more annoying is that this must be profitable. Who is falling for these scams where the name field in a transactional email says "call this number"

Internally, we use the term "CarPlay" for this. I would never buy a car that doesn't have Apple CarPlay. We try to identify the CarPlay features (the things our competitors have that ICPs can't live with out). Once we've satisfied that, then we ignore the competition and do our own thing.

Jason Cohen@asmartbear.com · 2mo ago

Should you ignore the competition? Yes, in that you need your own voice & product, leveraging your strengths and goals, not someone else's. No, too, because customers compare you to the competition. How you position and compare, is relevant. Ignorance is not strength.

Life in 2026: - Get an email from my bank saying my account is suspended, and to call them - Call and spend ~5 minutes navigating worthless AI support - Finally talk to a person who tells me they don't know why it's on hold, the algorithm just decided it. There's nothing to be done for now.

My least favorite type of post is when someone is like "oh god, no, please no" with no additional context and then it turns out they were just watching a baseball game and didnt like the outcome

Really doesn't feel like we're living in the future when a Stripe-issued credit card expires and every SaaS company I use has Stripe send me an email asking me to go to Stripe to tell them about my new credit card that I got from Stripe.

I love Mercury as a simple online business bank, but every new feature they add is so wildly useless to me. It's obvious they're targeting a totally different type of customer (high-growth VC-backed startup) and I just happen to like a small part of what that other customer wants.

This is exactly how I feel. If you'd told me I'd be able to work for myself and make an upper middle class income for 14+ years, I'd have been thrilled. And also, looking back, it seems like we could have achieved so much more. I guess it's healthy to feel a balance of content + ambitious?

Benedikt Deicke@benediktdeicke.com · 2mo ago

We're in year 9 of working on @userlist.com and some days it feels like we're way behind on our goals and ambitions 🥵 However, if you'd told my old self from back when we're started about where we're at now, my mind would've be blown 🤯

It's 2026. The most powerful technology ever is in our hands. Getting recommended by AI tools represents the next great frontier in marketing. And the key to it all, distilled but very true, remains getting listed in a bunch of crappy listicles. What a time to be alive.

It's probably just in my head, but it seems like Claude is acting really dumb today

The biggest benefit of airpods is that I now have headphones with me at all times. Wired headphones constantly got tangled in my pocket, so I only brought them with me if I knew I'd need them.

lauren@lauren.rotatingsandwiches.com · 3mo ago

i don't miss the headphone jack because wireless headphones are great and i don't miss earbuds getting violently ripped out of my head several times a day because i have the motor controls of a violently drunk chimpanzee

Something that's been on my mind: Is Meta's business actually hurt by the account takeover thing? Will anyone stop using them because of that? Downtime hurts because it means fewer page views, so less revenue. But do bugs, security vulns, etc. actually hurt monopolies? Do they need to even try?

Gergely Orosz@gergely.pragmaticengineer.com · 3mo ago

Meta had a SEV-0 outage today… less than two weeks after Meta’s most embarrassing undetected-for-too-long account takeover (also an outage) It’s impossible to unsee Meta pushing AI for code + reviews and the end result being more massive outages vs before They are connected

Two spam problems I really wish email providers solved: - I unsubscribe from a mailing list, but then they email me again 6 months later. I don't remember whether I unsubscribed already. - I get a terrible marketing email from, like, my bank. It's spam, but other email from that sender is important

We're rolling out a new code review flow: 1. Submit PR. 2. Run a local Claude Code skill to do a review. This pushes comments to the PR. 3. A github action runs a lighter review (since it's api billing) to flag any area that needs human review. 4. If anything is flagged, a senior dev reviews it.

Claude Code planning mode seems to miss the mark to me. It's always soooo eager to write the plan immediately. It's almost impossible to get it to truly brainstorm and spec out the project first. I guess you're just supposed to do that in a separate discussion first?

People underestimate how hard it is to get trustworthy self-reported data. For example, people say "ask users where they heard about you on the signup form" thinking that the answers will be correct. Here's a feedback form we just got. Note: Nothing about our software relates to phone calls.

Let me manifest this: I want AI's long-term equilibrium to be that it's super useful for code reviews and security audits, but too expensive / too dumb to meaningfully replace the need for human coders on a large, mature code base. That's the sweet spot.

We're entering the next phase of AI: When we adopt AI into a workflow, we're considering how it easy it will be to swap agents/models to go with whoever is offering the best subsidies at that moment. I want to both (a) get as much out of AI as possible and (b) drive their margins into the ground