Gobi
@gobiraj.bsky.social
Product Manager who questions the obvious | Building lovable products ❤️ | Sharing insights with a dash of humor | https://gobiraj.substack.com/
Pinned
Gobi
@gobiraj.bsky.social
· Nov 12
7 Rules of Power: The Career Advice No One Gives You
gobiraj.substack.com...
#ProductManagement #Power #Career
gobiraj.substack.com...
#ProductManagement #Power #Career
All AI-generated apps look the same because prompts default to the same patterns.
Fix: Be explicit about design elements. Color schemes, corner radii, font weights.
Or feed it screenshots of designs you actually want.
Generic prompts = generic output.
#AI
Fix: Be explicit about design elements. Color schemes, corner radii, font weights.
Or feed it screenshots of designs you actually want.
Generic prompts = generic output.
#AI
January 24, 2026 at 12:56 AM
All AI-generated apps look the same because prompts default to the same patterns.
Fix: Be explicit about design elements. Color schemes, corner radii, font weights.
Or feed it screenshots of designs you actually want.
Generic prompts = generic output.
#AI
Fix: Be explicit about design elements. Color schemes, corner radii, font weights.
Or feed it screenshots of designs you actually want.
Generic prompts = generic output.
#AI
Microservices trade-off nobody warns you about:
Monolith: transactions are easy, deployments are scary
Microservices: deployments are easy, transactions are heavy
Know what you're signing up for. Distributed systems have distributed problems.
#ProductManagement
Monolith: transactions are easy, deployments are scary
Microservices: deployments are easy, transactions are heavy
Know what you're signing up for. Distributed systems have distributed problems.
#ProductManagement
January 23, 2026 at 6:21 PM
Microservices trade-off nobody warns you about:
Monolith: transactions are easy, deployments are scary
Microservices: deployments are easy, transactions are heavy
Know what you're signing up for. Distributed systems have distributed problems.
#ProductManagement
Monolith: transactions are easy, deployments are scary
Microservices: deployments are easy, transactions are heavy
Know what you're signing up for. Distributed systems have distributed problems.
#ProductManagement
Analytics is everyone's priority until quarterly planning.
Then it's "we'll add tracking later."
Later never comes.
Build analytics into your definition of done. Not as a nice-to-have
#ProductManagement
Then it's "we'll add tracking later."
Later never comes.
Build analytics into your definition of done. Not as a nice-to-have
#ProductManagement
January 22, 2026 at 9:23 PM
Analytics is everyone's priority until quarterly planning.
Then it's "we'll add tracking later."
Later never comes.
Build analytics into your definition of done. Not as a nice-to-have
#ProductManagement
Then it's "we'll add tracking later."
Later never comes.
Build analytics into your definition of done. Not as a nice-to-have
#ProductManagement
PMs don't need to write code.
You need to understand enough to:
• Spot technically impractical ideas before proposing them
• Have meaningful discussions about timelines
• Ask good questions in architecture reviews
#ProductManagement
You need to understand enough to:
• Spot technically impractical ideas before proposing them
• Have meaningful discussions about timelines
• Ask good questions in architecture reviews
#ProductManagement
January 21, 2026 at 5:42 PM
PMs don't need to write code.
You need to understand enough to:
• Spot technically impractical ideas before proposing them
• Have meaningful discussions about timelines
• Ask good questions in architecture reviews
#ProductManagement
You need to understand enough to:
• Spot technically impractical ideas before proposing them
• Have meaningful discussions about timelines
• Ask good questions in architecture reviews
#ProductManagement
Stop trying to "learn AI" in the abstract.
Pick one real problem in your workflow.
Use one AI tool to solve it.
Do this for 30 days straight.
You'll learn more than any course could teach you.
#ProductManagement #AI
Pick one real problem in your workflow.
Use one AI tool to solve it.
Do this for 30 days straight.
You'll learn more than any course could teach you.
#ProductManagement #AI
January 20, 2026 at 3:27 PM
Stop trying to "learn AI" in the abstract.
Pick one real problem in your workflow.
Use one AI tool to solve it.
Do this for 30 days straight.
You'll learn more than any course could teach you.
#ProductManagement #AI
Pick one real problem in your workflow.
Use one AI tool to solve it.
Do this for 30 days straight.
You'll learn more than any course could teach you.
#ProductManagement #AI
Wild how vibe coding reversed expectations.
Building the app: easy
Maintaining it: nightmare
AI generates functional code but lacks architectural judgment. Every fix feels like starting from scratch.
Fast to ship ≠ easy to sustain.
#ProductManagement #AI
Building the app: easy
Maintaining it: nightmare
AI generates functional code but lacks architectural judgment. Every fix feels like starting from scratch.
Fast to ship ≠ easy to sustain.
#ProductManagement #AI
January 19, 2026 at 3:31 PM
Wild how vibe coding reversed expectations.
Building the app: easy
Maintaining it: nightmare
AI generates functional code but lacks architectural judgment. Every fix feels like starting from scratch.
Fast to ship ≠ easy to sustain.
#ProductManagement #AI
Building the app: easy
Maintaining it: nightmare
AI generates functional code but lacks architectural judgment. Every fix feels like starting from scratch.
Fast to ship ≠ easy to sustain.
#ProductManagement #AI
"We get to sprinkle AI on top!"
No. Connecting to OpenAI's APIs won't cut it in 2026.
AI-native means flipping the stack: AI Data Layer becomes foundation, not frosting.
That's not enhancement. That's a different architecture.
No. Connecting to OpenAI's APIs won't cut it in 2026.
AI-native means flipping the stack: AI Data Layer becomes foundation, not frosting.
That's not enhancement. That's a different architecture.
10 Things To Do Right Now to Become AI Native with Filevine's CEO & Founder
10 Things To Do Right Now to Become AI Native with Filevine's CEO & Founder
Ryan Anderson, Co-Founder and CEO of Filevine, shares the playbook for how his legal tech company successfully transitioned from a traditional SaaS business to an AI-native company, now generating more new revenue from AI p
www.youtube.com
January 19, 2026 at 2:19 PM
"We get to sprinkle AI on top!"
No. Connecting to OpenAI's APIs won't cut it in 2026.
AI-native means flipping the stack: AI Data Layer becomes foundation, not frosting.
That's not enhancement. That's a different architecture.
No. Connecting to OpenAI's APIs won't cut it in 2026.
AI-native means flipping the stack: AI Data Layer becomes foundation, not frosting.
That's not enhancement. That's a different architecture.
People’s IQs seem to double as soon as you give them responsibility and indicate that you trust them
January 18, 2026 at 2:40 AM
People’s IQs seem to double as soon as you give them responsibility and indicate that you trust them
As AI handles more execution-focused tasks, the demand for PMs will pivot toward strategic thinking and deep problem definition. Your value lies in validating why something should be built, not just managing its delivery.
#FutureOfWork #ProductStrategy
#FutureOfWork #ProductStrategy
January 16, 2026 at 11:01 PM
As AI handles more execution-focused tasks, the demand for PMs will pivot toward strategic thinking and deep problem definition. Your value lies in validating why something should be built, not just managing its delivery.
#FutureOfWork #ProductStrategy
#FutureOfWork #ProductStrategy
⚠️ AI is great at reinforcing your pre-existing assumptions. To build truly user-centric products, you must validate AI-generated insights with human testing to catch what the algorithm overlooks.
#ProductStrategy #UserTesting
#ProductStrategy #UserTesting
January 15, 2026 at 3:56 PM
⚠️ AI is great at reinforcing your pre-existing assumptions. To build truly user-centric products, you must validate AI-generated insights with human testing to catch what the algorithm overlooks.
#ProductStrategy #UserTesting
#ProductStrategy #UserTesting
A project can be flawlessly executed and still fail
You delivered on time, on budget, on spec. But the market moved. Or the problem wasn't real enough.
Execution excellence ≠ product success
Question: Should this continue? Not just: Are we doing this well?
#ProductManagement
You delivered on time, on budget, on spec. But the market moved. Or the problem wasn't real enough.
Execution excellence ≠ product success
Question: Should this continue? Not just: Are we doing this well?
#ProductManagement
January 13, 2026 at 8:34 PM
A project can be flawlessly executed and still fail
You delivered on time, on budget, on spec. But the market moved. Or the problem wasn't real enough.
Execution excellence ≠ product success
Question: Should this continue? Not just: Are we doing this well?
#ProductManagement
You delivered on time, on budget, on spec. But the market moved. Or the problem wasn't real enough.
Execution excellence ≠ product success
Question: Should this continue? Not just: Are we doing this well?
#ProductManagement
Most AI pilots fail because we underestimate what "working" actually requires
The AI has to do three things perfectly:
✓ Understand the intent
✓ Know the answer
✓ Do the work
Miss either one, and you're paying for a chatbot that routes work back to humans
#AI
The AI has to do three things perfectly:
✓ Understand the intent
✓ Know the answer
✓ Do the work
Miss either one, and you're paying for a chatbot that routes work back to humans
#AI
January 13, 2026 at 2:34 PM
Most AI pilots fail because we underestimate what "working" actually requires
The AI has to do three things perfectly:
✓ Understand the intent
✓ Know the answer
✓ Do the work
Miss either one, and you're paying for a chatbot that routes work back to humans
#AI
The AI has to do three things perfectly:
✓ Understand the intent
✓ Know the answer
✓ Do the work
Miss either one, and you're paying for a chatbot that routes work back to humans
#AI
Alignment isn't just a "task", it's the job. Expect to spend more time building consensus across stakeholders than on actual execution.
The "alignment tax" only grows as your company does.
#ProductManagement
The "alignment tax" only grows as your company does.
#ProductManagement
January 12, 2026 at 9:53 PM
Alignment isn't just a "task", it's the job. Expect to spend more time building consensus across stakeholders than on actual execution.
The "alignment tax" only grows as your company does.
#ProductManagement
The "alignment tax" only grows as your company does.
#ProductManagement
Before you "fix" a broken process, understand why it exists.
Chesterton's Fence: Don't tear down what you don't understand.
That sprint chaos might be protecting the team from something worse upstream.
Learn, then lead change.
#ProductManagement
Chesterton's Fence: Don't tear down what you don't understand.
That sprint chaos might be protecting the team from something worse upstream.
Learn, then lead change.
#ProductManagement
January 12, 2026 at 3:28 PM
Before you "fix" a broken process, understand why it exists.
Chesterton's Fence: Don't tear down what you don't understand.
That sprint chaos might be protecting the team from something worse upstream.
Learn, then lead change.
#ProductManagement
Chesterton's Fence: Don't tear down what you don't understand.
That sprint chaos might be protecting the team from something worse upstream.
Learn, then lead change.
#ProductManagement
"When one Waymo learns, they all learn."
We spend decades training individual humans.
Humans can’t download experience, but machines can sync models.
This is why AI out-scales us.
#AI
We spend decades training individual humans.
Humans can’t download experience, but machines can sync models.
This is why AI out-scales us.
#AI
January 10, 2026 at 3:04 PM
"When one Waymo learns, they all learn."
We spend decades training individual humans.
Humans can’t download experience, but machines can sync models.
This is why AI out-scales us.
#AI
We spend decades training individual humans.
Humans can’t download experience, but machines can sync models.
This is why AI out-scales us.
#AI
Most people sprinkle AI on top of a product like salt.
Better way: Break your JTBD down into granular skills. If an LLM or agent can perform a specific skill better/faster than a human, that’s where you build.
Solve for the skill, not the buzzword.
#AI
Better way: Break your JTBD down into granular skills. If an LLM or agent can perform a specific skill better/faster than a human, that’s where you build.
Solve for the skill, not the buzzword.
#AI
January 10, 2026 at 12:06 AM
Most people sprinkle AI on top of a product like salt.
Better way: Break your JTBD down into granular skills. If an LLM or agent can perform a specific skill better/faster than a human, that’s where you build.
Solve for the skill, not the buzzword.
#AI
Better way: Break your JTBD down into granular skills. If an LLM or agent can perform a specific skill better/faster than a human, that’s where you build.
Solve for the skill, not the buzzword.
#AI
Stop making your LLMs start from scratch every time
The most efficient design pattern emerging right now is Recursive Tool Building: having an LLM write deterministic code that it (or future sessions) can call to solve repetitive tasks
How to build it 👇
1/4
The most efficient design pattern emerging right now is Recursive Tool Building: having an LLM write deterministic code that it (or future sessions) can call to solve repetitive tasks
How to build it 👇
1/4
January 8, 2026 at 11:39 PM
Stop making your LLMs start from scratch every time
The most efficient design pattern emerging right now is Recursive Tool Building: having an LLM write deterministic code that it (or future sessions) can call to solve repetitive tasks
How to build it 👇
1/4
The most efficient design pattern emerging right now is Recursive Tool Building: having an LLM write deterministic code that it (or future sessions) can call to solve repetitive tasks
How to build it 👇
1/4
Most PRDs fail because they’re either a novel or a sticky note.
To ship high-quality features without the friction, you only need three things:
1️⃣ The Problem (Why are we doing this?)
2️⃣ The Outcome (What does success look like?)
3️⃣ The Prototype (How does it feel?)
To ship high-quality features without the friction, you only need three things:
1️⃣ The Problem (Why are we doing this?)
2️⃣ The Outcome (What does success look like?)
3️⃣ The Prototype (How does it feel?)
January 8, 2026 at 6:25 PM
Most PRDs fail because they’re either a novel or a sticky note.
To ship high-quality features without the friction, you only need three things:
1️⃣ The Problem (Why are we doing this?)
2️⃣ The Outcome (What does success look like?)
3️⃣ The Prototype (How does it feel?)
To ship high-quality features without the friction, you only need three things:
1️⃣ The Problem (Why are we doing this?)
2️⃣ The Outcome (What does success look like?)
3️⃣ The Prototype (How does it feel?)
Malleable software: tools that fit you like a good leather shoe, not the other way around.
We've spent decades bending to our software's constraints. Now AI lets us evolve our tools as we use them.
We're moving from "learn the tool" to "the tool learns you."
#AI
We've spent decades bending to our software's constraints. Now AI lets us evolve our tools as we use them.
We're moving from "learn the tool" to "the tool learns you."
#AI
January 8, 2026 at 2:22 PM
Malleable software: tools that fit you like a good leather shoe, not the other way around.
We've spent decades bending to our software's constraints. Now AI lets us evolve our tools as we use them.
We're moving from "learn the tool" to "the tool learns you."
#AI
We've spent decades bending to our software's constraints. Now AI lets us evolve our tools as we use them.
We're moving from "learn the tool" to "the tool learns you."
#AI
Informal weekly stakeholder updates > Formal monthly reports
Build trust through frequent, casual touchpoints.
By the time big decisions come, you've already earned the buy-in.
"Dig the well before you are thirsty"
#ProductManagement
Build trust through frequent, casual touchpoints.
By the time big decisions come, you've already earned the buy-in.
"Dig the well before you are thirsty"
#ProductManagement
January 7, 2026 at 9:14 PM
Informal weekly stakeholder updates > Formal monthly reports
Build trust through frequent, casual touchpoints.
By the time big decisions come, you've already earned the buy-in.
"Dig the well before you are thirsty"
#ProductManagement
Build trust through frequent, casual touchpoints.
By the time big decisions come, you've already earned the buy-in.
"Dig the well before you are thirsty"
#ProductManagement
Traditional acceptance criteria break with AI.
Nondeterministic outputs need continuous evals, not one-time sign-offs.
Your QA process is now an ongoing process, not a gate.
#ProductManagement #AI
Nondeterministic outputs need continuous evals, not one-time sign-offs.
Your QA process is now an ongoing process, not a gate.
#ProductManagement #AI
January 7, 2026 at 2:08 AM
Traditional acceptance criteria break with AI.
Nondeterministic outputs need continuous evals, not one-time sign-offs.
Your QA process is now an ongoing process, not a gate.
#ProductManagement #AI
Nondeterministic outputs need continuous evals, not one-time sign-offs.
Your QA process is now an ongoing process, not a gate.
#ProductManagement #AI
PMs: If it takes longer to do manually than to describe to an AI, automate it. Use AI to handle the small stuff so you can focus on moving the needle.
Read more on the hidden cost of not automating: gobiraj.substack.com...
#AI
Read more on the hidden cost of not automating: gobiraj.substack.com...
#AI
The Hidden Cost of Not Automating: When Tiny Tasks Derail Big Thinking
AI automation is the new superpower for forward-thinking PMs
gobiraj.substack.com
January 6, 2026 at 1:39 PM
PMs: If it takes longer to do manually than to describe to an AI, automate it. Use AI to handle the small stuff so you can focus on moving the needle.
Read more on the hidden cost of not automating: gobiraj.substack.com...
#AI
Read more on the hidden cost of not automating: gobiraj.substack.com...
#AI
Visual‑language‑action stack = Full AI pipeline that lets a humanoid robot see, understand instructions, and then move its body to do a task
bostondynamics.com/b...
#AI
bostondynamics.com/b...
#AI
Boston Dynamics & Google DeepMind Form New AI Partnership to Bring Foundational Intelligence to Humanoid Robots | Boston Dynamics
A new robotics partnership aims to combine Boston Dynamics’ athletic intelligence with Google DeepMind’s foundational capabilities
bostondynamics.com
January 6, 2026 at 6:13 AM
Visual‑language‑action stack = Full AI pipeline that lets a humanoid robot see, understand instructions, and then move its body to do a task
bostondynamics.com/b...
#AI
bostondynamics.com/b...
#AI
Drowning in meetings: a simple formula for effective meetings
No agenda = No meeting
No clear outcome = No meeting
No one leaves with action items = Wasted meeting
That's it. That's the formula.
#ProductManagement
No agenda = No meeting
No clear outcome = No meeting
No one leaves with action items = Wasted meeting
That's it. That's the formula.
#ProductManagement
January 6, 2026 at 12:03 AM
Drowning in meetings: a simple formula for effective meetings
No agenda = No meeting
No clear outcome = No meeting
No one leaves with action items = Wasted meeting
That's it. That's the formula.
#ProductManagement
No agenda = No meeting
No clear outcome = No meeting
No one leaves with action items = Wasted meeting
That's it. That's the formula.
#ProductManagement