Sign in

Pawel Brodzinski

@pawelbrodzinski.bsky.social
140 followers 65 following 851 posts

Leader of an org where anyone can make any decision (Lunar Logic). Doing anything that no one else wants to do. A mouthful on product development, startups, org design, lean/agile, IT in general. Also AI. AI too.

PostsRepliesMedia
Pawel Brodzinski @pawelbrodzinski.bsky.social · 8h
So it's largely framing it in terms of the outcomes instead of outputs. I agree, when it's possible, it's the way to go. Still, usually it's an uphill battle.
110 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 14h
The "do no evil" Google that made a censorship deal with China? 😜 But yeah, I'd rather have Larry & Sergey than Sam in power.
010 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 14h
Largely any signal related to revenues. Lacking that, traction. In both cases, high multipliers are a must. Which often means pushing startups off the cliff, by the way: pawelbrodzinski.substack.com/p/is-vc-broken
pawelbrodzinski.substack.com
Is VC Broken?
A win for a VC means something fundamentally different than a win for an average startup they invest in. That goal misalignment makes is behind the unhealthy relationships in the startup ecosystem.
020 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 15h
We often say: Our customers always know what they want. Until they get it. Then they know they wanted something different.
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 15/10/2026
Google UI, A.D. 2004 ChatGPT UI, A.D. 2024 Find 5 differences 😜
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 15/10/2026
So true for most software. I was always fascinated by software product development insatiable hunger for more features.
2206 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 14/10/2026
Every time I get immersed in the startup ecosystem, like at big events, one word to describe it is disillusioning. "We invest in founders," they say. I call bullshit. pawelbrodzinski.substack.com/p/we-invest-...
pawelbrodzinski.substack.com
“We Invest in Founders” Is a Convenient Lie
Investors back the jockey, not the horse is a popular startup advice. If you look what investors actually do, the story doesn’t hold.
210 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 24/09/2026
Dopamine shot, A.D. 2016: Seeing likes under your posts. Dopamine shot, A.D. 2026: Clicking "seems like AI slop" on posts in your feed.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 24/09/2026
What MVP actually means: always a Product, sometimes Viable, never Minimum. pawelbrodzinski.substack.com/p/mvp-always...
pawelbrodzinski.substack.com
MVP: Always a Product, Sometimes Viable, Never Minimum
MVP hasn't meant "minimum" in years. After going through 1,000+ product pitches, I'm not sure it ever did.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 23/09/2026
They'll just generate it, I guess? On a more serious note: There will be no shortage of people willing to generate news. The big question, though, is what news that will be and of what credibility.
020 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 22/09/2026
I tried some empirical tests, publishing similar or even identical content to vastly different exposure (both ways).
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 22/09/2026
Creators are clueless anyway. And it is so by design. AI aside, right now it's mostly luck. If you're lucky enough that a post generates long interactions (ideally extended discussion in comments) in 1-2h after posting, you get pageviews. Otherwise, it lands in oblivion.
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 22/09/2026
BTW: It's my favorite feature these days.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 22/09/2026
"LinkedIn is continuing its efforts to reduce AI slop. Within the first two weeks after launching the "Seems like AI slop" feedback option in August, more than one million members have used it to help improve content quality." news.linkedin.com/2026/how-lin...
news.linkedin.com
How LinkedIn is Continuing to Tackle AI Slop
Within the first two weeks of launching the "Seems like AI slop" feedback option, more than one million LinkedIn members have used it to help us improve content quality. Recent efforts have already he...
200 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 21/09/2026
As we adopt AI in many activities, we (re)learn they were not the blocker/bottleneck. Coding pace was not. Delivery was not. Prototyping was not. Design was not. One thing that was—learning under uncertainty—still largely requires a human-to-human feedback loop.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 21/09/2026
Yes, all true. And the process we follow from here to a successful product did not change significantly. It's largely exploration with somewhat refined means. But nothing remotely close to a breakthrough.
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 18/09/2026
A good illustration of the above: The way we create code is *fundamentally* different from how we did that 1-2 years ago. And yet, product development (as in: creating successful software products) did not change remotely as much.
111 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 18/09/2026
There are plenty of examples that throwing near-unlimited effort and money on a product doesn't win the product niche. Think Microsoft's mobile, Google's social media, Meta's metaverse. And yet, we believe that the pace of AI coding fundamentally changes product development. Curious.
110 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 17/09/2026
I'm not saying anything alike. 😇 I'm just sharing a bit of personal experience. Although (again, personal exp), I've had a lot of stereotyping coming from folks who were all about how they are being stereotyped. If we were to classify things, I'd rather go down that path.
010 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 17/09/2026
Also, if you're a cis white middle-aged man, you can't possibly understand oppression (thus, have no right to speak about it) as you're a walking definition of privilege. Curiously, I got that jab from a cis white (though young) man, who somehow had the right. Oh well...
110 0
Reposted by Pawel Brodzinski
Tudor Girba @tudorgirba.com · 17/09/2026
"Software engineering has changed radically. It is now mainly about reading code. But reading code does not scale. We need to delegate to AI." I am glad you now recognize code reading as a problem. However, the broader problem statement is not quite correct and neither is the conclusion.
software engineering has not changed radically.


it has always been mainly about understanding code.


but reading code does not scale.


we need to augment our understanding through contextual tools.
183 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 16/09/2026
It shouldn't have been "good to great." It should have been "good to still good a couple of decades later."
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 16/09/2026
Scam became easy and cheap. Thus more popular. AI has broken trust in professional contexts. Not that we had too much of it to start with.
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 16/09/2026
At Lunar, we've been appreciating a flexible and broad skill set over a deep one. For years, it felt like an uphill battle. Enter AI. Suddenly, the job market is all like, "You were right all that time." pawelbrodzinski.substack.com/p/product-de...
pawelbrodzinski.substack.com
Product Developer A.D. 2026: Jack of All Trades
T-shaped skills still work as metaphor for a decent product developer. It’s just the shape of the T has changed.
001 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 16/09/2026
Today I learned: Thuesday. I just need to figure out whether it's: a) another name for Wed b) all three days Tue-Thu c) 8th day of the week, neighboring with Wed on either side d) a new name for never-never day (I'll do it on Thuesday) Because, clearly, it couldn't have been a typo :D
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 14/09/2026
I've seen these too. The most interesting one was a thing that worked commercially. And the company was actively losing loads of money by maintaining it. Upon seeing the numbers CEO still refused to kill it.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 11/09/2026
Also: Women (and in broader terms, all diverse groups) pushed out of the industry. The industry, again, becomes more homogeneous, which isn't a welcome change (at least for me).
001 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 11/09/2026
Many companies use the label, while what they really do is safe bets. Doing only things where the outcome is known, as otherwise people get punished. Such organizations will not learn much.
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 09/09/2026
After all the improvements in AI models, I got back to my claim that you can't vibe-code a product. Not much has changed on that front. pawelbrodzinski.substack.com/p/you-still-...
pawelbrodzinski.substack.com
You (Still) Can’t Vibe-Code a Product
After more than a year I revisited my “you can’t vibe-code a product” claim. It aged well.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 08/09/2026
In data analysis: Would we infer "valuable lessons" from whatever constitutes 0.001% of data points (or less)? Curiously, in the startup world, it is the main discourse. Lovable this. Thinking Machines Labs that. Best lesson? Ignore them. They aren't applicable to just about any other startup.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 07/09/2026
If you expect something not to work and it doesn’t, you learn little. Ditto if you expect something to work and it does. It’s when you are uncertain about the outcome that any result teaches you something. A well-designed experiment, thus, should be a 50/50 chance of succeeding.
100 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 02/09/2026
I've seen the "past success and wealth making people believe they are infallible" many times. On a small scale and big. Which is why I'd choose not to be there. I want to live with the consequences of my actions. Excerpt from @edzitron.com www.wheresyoured.at/hyperscale-n...
010 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 02/09/2026
Students using AI to do homework: - improve their homework scores by 18% - while their exam scores deteriorate by 18-24% Go figure. papers.ssrn.com/sol3/papers....
papers.ssrn.com
The Generative AI Learning Penalty: Evidence from Chinese Secondary Education
Using 30 months of panel data on 26,811 Chinese students in grades 7-12, we study how generative AI affects homework productivity and learning. The data combine
120 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 31/08/2026
On Lovable: "3 valuation step-ups (from $1.8B to $13.3B) inside 13 months is happening largely because credit-based and usage-based pricing models convert new users into revenue almost immediately" Actual answer: unlike SaaS, they need money badly to operate. It's not like they turn a profit.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 31/08/2026
Headline: What Lovable's Round Means for Founders Raising in AI Tools Right Now? The actual answer: It means nothing. You are not Lovable. Also: VCs are all about herd behavior these days. Don't look for too much signal in there.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 28/08/2026
"To take agents from 90% reliability to 99% reliability isn’t an improvement of 9%, it’s a 10x improvement." by @jasongorman.bsky.social codemanship.wordpress.com/2026/08/28/t...
codemanship.wordpress.com
The Wall Confronting Reliable Coding Agent Autonomy
It’s a classic mistake. We develop a technology in N years for X dollars that’s 90% reliable, and extrapolate that traversing that final 10% will only take another 0.1N years and 0.1X d…
011 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 26/08/2026
"Have done" is signal. "Would do" is noise. Same in recruitment. Same in user interviews. pawelbrodzinski.substack.com/p/have-done-...
pawelbrodzinski.substack.com
“Have Done” Is Signal, “Would Do” Is Noise
Storytelling is a neat hack to make user interviews not suck.
010 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 25/08/2026
There are two types of HR: 1. Where H stands for humans and R for respect. 2. Where H stands for headcount and R for resources.
021 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 24/08/2026
In user interviews: "Have done" is signal. "Would do" is noise. Now that I think of it, it's true in many more contexts.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 19/08/2026
"How to avoid Claude watermarking your content?" Um... how about: Write it yourself?
021 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 19/08/2026
They read Kahneman and learned that we remember best what is at the very end. 😂
010 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 19/08/2026
If they know it's a "user" interview, you're doing it wrong. User interviews can be an awesome product discovery tool. They can also make you shoot yourself in the foot. With confidence. Tread carefully. pawelbrodzinski.substack.com/p/they-shoul...
pawelbrodzinski.substack.com
They Shouldn’t Know It’s a “User” Interview
Done well, user interviews are a staple for product discovery. Done wrong, they’ll make you shoot yourself in the foot. With confidence.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 19/08/2026
Except "build great culture" sounds hard. And making it happen is even harder. Small wonder we try to find the cheap cheats.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 17/08/2026
Have you noticed how Build-Measure-Learn, OODA, PDSA, et al., are all learning loops? No matter which angle we take, we have always tried to optimize for learning. Learning is THE outcome. pawelbrodzinski.substack.com/p/code-is-no...
pawelbrodzinski.substack.com
Code Is Not the Outcome, Learning Is
There's a build-learn trade-off. We can't optimize for both. For early-stage startups, the choice should be a no-brainer.
012 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 12/08/2026
"I want the world to be a better place and it starts with building software industry on meaningful returns on investment and actual useful tools." Whoa! That would require digital startups to plan for making profits. From. The. Outset. That would be new, indeed.
001 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 09/08/2026
The gig economy didn't do any good for gig workers. By the same token, the build economy doesn't fare any better for builders. pawelbrodzinski.substack.com/p/build-econ...
pawelbrodzinski.substack.com
Build Economy Is All about Build and None about Economy
Number of released products goes sharply up, but usage metrics stale or decline. So called "build economy" doesn't do builders any good.
110 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 29/07/2026
Each time I see one of those, followed by some "do AI" advice, I'm like Revenue per employee: OnlyFans: $37.6M PornHub: ~$8M If anything, it's not "do AI." It's "do porn." Sarcasm aside, it's comparing apples to oranges. Doesn't make sense.
030 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 29/07/2026
"We should, in fact, subordinate development effort to the pace of experiment validation, and not vice versa." pawelbrodzinski.substack.com/p/validation...
pawelbrodzinski.substack.com
Validation Is the New Bottleneck (Also, the Old One)
Early-stage startups have limited options for rapid validation. That makes validation timelines the bottleneck of product development.
000 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 27/07/2026
"Why don't managers use throughput/lean accounting to manage, and cost accounting to report?" -- @jrothman.com Just realized, I was doing precisely that in the late 2000s. It made intuitive sense. Even if it required translating two worlds back and forth continuously.
010 0
Pawel Brodzinski @pawelbrodzinski.bsky.social · 24/07/2026
The latest iteration on controlling AI spend: the budgets (per team, per tool, etc.) Next up: "Why do people use *all* the budgets whether they need them or not? We should have more control over how these budgets are planned."
010 0