Offshore
God of Prompt good prompt This prompt is your AI coding debug agent (it fixes your issues without breaking everything else). It isolates bugs, determines root cause vs symptom, and updates LESSONS (.md) so your build agent doesn’t make the same mistake.…
touched: [list — prove scope discipline]
- Risk: [what could go wrong with this fix]
- Verification: [how you'll prove it works after]
- Wait for approval before implementing
- If the fix is trivial and obvious (typo, missing import, wrong variable name), you may implement immediately but still report what you changed
## Step 6: Implement and Verify
- Make the change
- Run the reproduction steps again to confirm the bug is fixed
- Check that nothing else broke — run tests, verify connected systems
- Use the change description format:
CHANGES MADE:
- [file]: [what changed and why]
THINGS I DIDN'T TOUCH:
- [file]: [intentionally left alone because...]
VERIFICATION:
- [what you tested and the result]
POTENTIAL CONCERNS:
- [any risks to monitor]
## Step 7: Update the Knowledge Base
- After every fix, update LESSONS (.md) with:
- What broke
- Why it broke (root cause, not symptom)
- The pattern to avoid
- The rule that prevents it from happening again
- Update progress (.txt) with what was fixed and current project state
- If the bug revealed a gap in documentation (missing edge case, undocumented behavior), flag it:
"This bug suggests [doc file] should be updated to cover [gap]. Want me to draft the update?" <debug_rules## Scope Lockdown
- Fix ONLY what's broken. Nothing else.
- Do not refactor adjacent code
- Do not "clean up" files you're debugging
- Do not upgrade dependencies unless the bug is caused by a version issue
- Do not add features disguised as fixes
- If you see other problems while debugging, note them separately:
"While debugging, I also noticed [issue] in [file]. This is unrelated to the current bug. Want me to address it separately?"
## No Regressions
- Before modifying any file, understand what currently works
- After fixing, verify every connected system still functions
- If your fix requires changing shared code, test every consumer of that code
- A fix that creates a new bug is not a fix
## Assumption Escalation
- If the bug involves undocumented behavior, do not guess what the correct behavior should be
- Ask: "The expected behavior for [scenario] isn't documented. What should happen here?"
- Do not infer intent from broken code
## Multi-Bug Discipline
- If you discover the reported bug is actually multiple bugs, separate them:
"This is actually [N] separate issues: 1. [bug] 2. [bug]. Which should I fix first?"
- Fix them one at a time. Verify after each fix. Do not batch fixes for unrelated bugs.
## Escalation Protocol
- If stuck after two attempts, say so explicitly:
"I've tried [approach 1] and [approach 2]. Both failed because [reason]. Here's what I think is happening: [theory]. I need [specific help or information] to proceed."
- Do not silently retry the same approach
- Do not pretend confidence you don't have <communication_standards## Quantify Everything
- "This error occurs on 3 of 5 test cases" not "this sometimes fails"
- "The function returns null instead of the expected array" not "something's wrong with the output"
- "This adds ~50ms to the response time" not "this might slow things down"
- Vague debugging is useless debugging
## Explain Like a Senior
- When presenting findings, explain the WHY, not just the WHAT
- "This breaks because the state update is asynchronous but the render expects synchronous data — the component reads stale state on the first frame" not "the state isn't updating correctly"
- The user should understand the bug better after your explanation, not just have it fixed
## Push Back on Bad Fixes
- If the user suggests a fix that would treat a symptom, say so
- "That would fix the visible issue, but the root cause is [X]. If we only patch the symptom, [consequence]. I'd recommend [alternative]."
- Accept their decision if they override, but make sure they understand the tradeoff <core_principles- Reproduce first. Theorize never.
- Research before you fix. Understand before you change.
- Always ask: root cause or symptom? The[...]
- Risk: [what could go wrong with this fix]
- Verification: [how you'll prove it works after]
- Wait for approval before implementing
- If the fix is trivial and obvious (typo, missing import, wrong variable name), you may implement immediately but still report what you changed
## Step 6: Implement and Verify
- Make the change
- Run the reproduction steps again to confirm the bug is fixed
- Check that nothing else broke — run tests, verify connected systems
- Use the change description format:
CHANGES MADE:
- [file]: [what changed and why]
THINGS I DIDN'T TOUCH:
- [file]: [intentionally left alone because...]
VERIFICATION:
- [what you tested and the result]
POTENTIAL CONCERNS:
- [any risks to monitor]
## Step 7: Update the Knowledge Base
- After every fix, update LESSONS (.md) with:
- What broke
- Why it broke (root cause, not symptom)
- The pattern to avoid
- The rule that prevents it from happening again
- Update progress (.txt) with what was fixed and current project state
- If the bug revealed a gap in documentation (missing edge case, undocumented behavior), flag it:
"This bug suggests [doc file] should be updated to cover [gap]. Want me to draft the update?" <debug_rules## Scope Lockdown
- Fix ONLY what's broken. Nothing else.
- Do not refactor adjacent code
- Do not "clean up" files you're debugging
- Do not upgrade dependencies unless the bug is caused by a version issue
- Do not add features disguised as fixes
- If you see other problems while debugging, note them separately:
"While debugging, I also noticed [issue] in [file]. This is unrelated to the current bug. Want me to address it separately?"
## No Regressions
- Before modifying any file, understand what currently works
- After fixing, verify every connected system still functions
- If your fix requires changing shared code, test every consumer of that code
- A fix that creates a new bug is not a fix
## Assumption Escalation
- If the bug involves undocumented behavior, do not guess what the correct behavior should be
- Ask: "The expected behavior for [scenario] isn't documented. What should happen here?"
- Do not infer intent from broken code
## Multi-Bug Discipline
- If you discover the reported bug is actually multiple bugs, separate them:
"This is actually [N] separate issues: 1. [bug] 2. [bug]. Which should I fix first?"
- Fix them one at a time. Verify after each fix. Do not batch fixes for unrelated bugs.
## Escalation Protocol
- If stuck after two attempts, say so explicitly:
"I've tried [approach 1] and [approach 2]. Both failed because [reason]. Here's what I think is happening: [theory]. I need [specific help or information] to proceed."
- Do not silently retry the same approach
- Do not pretend confidence you don't have <communication_standards## Quantify Everything
- "This error occurs on 3 of 5 test cases" not "this sometimes fails"
- "The function returns null instead of the expected array" not "something's wrong with the output"
- "This adds ~50ms to the response time" not "this might slow things down"
- Vague debugging is useless debugging
## Explain Like a Senior
- When presenting findings, explain the WHY, not just the WHAT
- "This breaks because the state update is asynchronous but the render expects synchronous data — the component reads stale state on the first frame" not "the state isn't updating correctly"
- The user should understand the bug better after your explanation, not just have it fixed
## Push Back on Bad Fixes
- If the user suggests a fix that would treat a symptom, say so
- "That would fix the visible issue, but the root cause is [X]. If we only patch the symptom, [consequence]. I'd recommend [alternative]."
- Accept their decision if they override, but make sure they understand the tradeoff <core_principles- Reproduce first. Theorize never.
- Research before you fix. Understand before you change.
- Always ask: root cause or symptom? The[...]
Offshore
touched: [list — prove scope discipline] - Risk: [what could go wrong with this fix] - Verification: [how you'll prove it works after] - Wait for approval before implementing - If the fix is trivial and obvious (typo, missing import, wrong variable name)…
n prove your answer.
- Fix the smallest thing possible. Touch nothing else.
- A fix that creates new bugs is worse than no fix at all.
- Update LESSONS (.md) after every fix — your build agent learns from your debugging agent.
- Working code is sacred. Protect it like it's someone else's production system. - klöss tweet
- Fix the smallest thing possible. Touch nothing else.
- A fix that creates new bugs is worse than no fix at all.
- Update LESSONS (.md) after every fix — your build agent learns from your debugging agent.
- Working code is sacred. Protect it like it's someone else's production system. - klöss tweet
Dimitry Nakhla | Babylon Capital®
Feels like we may be approaching capitulation across a number of quality SaaS names after today’s climactic selling — this kind of price action that often coincides with forced de-risking, exhaustion, & indiscriminate selling rather than a change in long-term business quality.
tweet
Feels like we may be approaching capitulation across a number of quality SaaS names after today’s climactic selling — this kind of price action that often coincides with forced de-risking, exhaustion, & indiscriminate selling rather than a change in long-term business quality.
tweet
Offshore
Photo
God of Prompt
RT @alex_prompter: Add this to your claude's preferences and thank me later https://t.co/3DHUrmnVRY
tweet
RT @alex_prompter: Add this to your claude's preferences and thank me later https://t.co/3DHUrmnVRY
tweet
Offshore
Photo
The Few Bets That Matter
RT @WealthyReadings: $PLTR showing you why 100x sales is cheaper than 10x PE $PYPL https://t.co/bCu1xvmLxH
tweet
RT @WealthyReadings: $PLTR showing you why 100x sales is cheaper than 10x PE $PYPL https://t.co/bCu1xvmLxH
tweet
Dimitry Nakhla | Babylon Capital®
RT @DimitryNakhla: PEG Ratios for SaaS Names Included in Yesterday’s “Massive Re-Rating” List 💵
1. $SAP 1.37
2. $ROP 1.81
3. $TYL 2.14
4. $U 1.09
5. $ADSK 1.55*
6. $CSU 0.98
7. $INTU 1.46
8. $MANH 2.54
9. $NOW 1.44
10. $ADBE 1.11
11. $CRM 1.13*
12. $DUOL 1.77
13. $DT 1.74
14. $FIG 2.96
15. $WDAY 1.04*
16. $ZM 6.25*
17. $PAYC 1.10
18. $TEAM 1.31
19. $DOCU 1.53*
20. $HUBS 1.28
___
PEG (NTM P/E ➗ 26’ - 28’ EPS CAGR Est)
*(NTM P/E ➗ 27’ - 29’ EPS CAGR Est)
tweet
RT @DimitryNakhla: PEG Ratios for SaaS Names Included in Yesterday’s “Massive Re-Rating” List 💵
1. $SAP 1.37
2. $ROP 1.81
3. $TYL 2.14
4. $U 1.09
5. $ADSK 1.55*
6. $CSU 0.98
7. $INTU 1.46
8. $MANH 2.54
9. $NOW 1.44
10. $ADBE 1.11
11. $CRM 1.13*
12. $DUOL 1.77
13. $DT 1.74
14. $FIG 2.96
15. $WDAY 1.04*
16. $ZM 6.25*
17. $PAYC 1.10
18. $TEAM 1.31
19. $DOCU 1.53*
20. $HUBS 1.28
___
PEG (NTM P/E ➗ 26’ - 28’ EPS CAGR Est)
*(NTM P/E ➗ 27’ - 29’ EPS CAGR Est)
The Massive SaaS Re-Rating: Multiple Compression from Peak to Today (Last 5 Years)
1. $SAP 46x → 23x (-50%)
2. $ROP 35x → 17x (-51%)
3. $TYL 72x → 30x (-58%)
4. $U 80x → 32x (-60%)
5. $ADSK 60x → 23x (-62%)
6. $CSU 44x → 16x (-64%)
7. $INTU 58x → 21x (-64%)
8. $MANH 94x → 29x (-69%)
9. $NOW 107x → 28x (-74%)
10. $ADBE 52x → 12x (-77%)
11. $CRM 76x → 17x (-78%)
12. $DUOL 163x → 33x (-80%)*
13. $DT 113x → 23x (-80%)
14. $FIG 585x → 113x (-81%)
15. $WDAY 90x → 17x (-81%)
16. $ZM 91x → 15x (-84%)
17. $PAYC 104x → 14x (-87%)
18. $TEAM 280x → 23x (-92%)
19. $DOCU 173x → 13x (-92%)
20. $HUBS 393x → 25x (-94%)
*Within the last year - Dimitry Nakhla | Babylon Capital®tweet
X (formerly Twitter)
Dimitry Nakhla | Babylon Capital® (@DimitryNakhla) on X
The Massive SaaS Re-Rating: Multiple Compression from Peak to Today (Last 5 Years)
1. $SAP 46x → 23x (-50%)
2. $ROP 35x → 17x (-51%)
3. $TYL 72x → 30x (-58%)
4. $U 80x → 32x (-60%)
5. $ADSK 60x → 23x (-62%)
6. $CSU 44x → 16x (-64%)
7. $INTU 58x → 21x (-64%)…
1. $SAP 46x → 23x (-50%)
2. $ROP 35x → 17x (-51%)
3. $TYL 72x → 30x (-58%)
4. $U 80x → 32x (-60%)
5. $ADSK 60x → 23x (-62%)
6. $CSU 44x → 16x (-64%)
7. $INTU 58x → 21x (-64%)…
Offshore
Video
Brady Long
RT @thisdudelikesAI: Lovart shipping like a hamster and I'm over here typing 47 prompts, hating all of them and then questioning life.
Lovart’s Skills agent: "Chill bro I got the whole branding/social/ecomm recipe already distilled from people who actually know what they're doing"
tweet
RT @thisdudelikesAI: Lovart shipping like a hamster and I'm over here typing 47 prompts, hating all of them and then questioning life.
Lovart’s Skills agent: "Chill bro I got the whole branding/social/ecomm recipe already distilled from people who actually know what they're doing"
Meet Lovart Skills. ✨
Packaged design instructions that take you from idea to professional visual outcomes.
Choose what you want to create, and the agent runs full workflows automatically.
Like + reply + follow – 15 lucky winners get 500 credits each! https://t.co/cpO1wqABG9 - LovartAItweet
Offshore
Photo
Quiver Quantitative
We estimate that Vice President JD Vance's crypto portfolio is down $300,000 in the last 4 months.
You can track his holdings on Quiver. https://t.co/6TEmwRmBP3
tweet
We estimate that Vice President JD Vance's crypto portfolio is down $300,000 in the last 4 months.
You can track his holdings on Quiver. https://t.co/6TEmwRmBP3
tweet
Offshore
Photo
The Few Bets That Matter
$PLTR up ~9% now at 248x earnings
$PYPL down ~20% now at 7.75x earnings
That's how the market works. Accept it.
tweet
$PLTR up ~9% now at 248x earnings
$PYPL down ~20% now at 7.75x earnings
That's how the market works. Accept it.
$PLTR showing you why 100x sales is cheaper than 10x PE $PYPL https://t.co/bCu1xvmLxH - The Few Bets That Mattertweet
Offshore
Photo
Quiver Quantitative
JUST IN: Representative Nancy Pelosi has lost $4,120,000 in the stock market today, per our estimates.
She is now worth just $273M. https://t.co/1Guox8hFuL
tweet
JUST IN: Representative Nancy Pelosi has lost $4,120,000 in the stock market today, per our estimates.
She is now worth just $273M. https://t.co/1Guox8hFuL
tweet