@itsalexzajac.bsky.social
If you enjoyed this thread:
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
Subscribe | Hungry Minds
Get smarter about Software and AI in 5 minutes. Save 50+ hours/week with deep dives, trends and tools hand-picked from 100+ sources. Join 50K+ engineers from big tech to startups for 1 free email every Monday.
www.hungryminds.dev
November 13, 2025 at 4:32 PM
If you enjoyed this thread:
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
Did you test out Gamma?
November 13, 2025 at 4:32 PM
Did you test out Gamma?
This won't replace a design team for your Series A deck.
But for technical demos, sprint reviews, and architecture docs?
→ It's faster
→ More maintainable
→ Better than what most of us were making anyway
But for technical demos, sprint reviews, and architecture docs?
→ It's faster
→ More maintainable
→ Better than what most of us were making anyway
November 13, 2025 at 4:32 PM
This won't replace a design team for your Series A deck.
But for technical demos, sprint reviews, and architecture docs?
→ It's faster
→ More maintainable
→ Better than what most of us were making anyway
But for technical demos, sprint reviews, and architecture docs?
→ It's faster
→ More maintainable
→ Better than what most of us were making anyway
After:
→ Write specs in Notion
→ Hit the Gamma API endpoint
→ Get the presentation in 3 minutes
→ Share live link that updates with source docs
Time on slides: 15 minutes, not 7 hours.
→ Write specs in Notion
→ Hit the Gamma API endpoint
→ Get the presentation in 3 minutes
→ Share live link that updates with source docs
Time on slides: 15 minutes, not 7 hours.
November 13, 2025 at 4:32 PM
After:
→ Write specs in Notion
→ Hit the Gamma API endpoint
→ Get the presentation in 3 minutes
→ Share live link that updates with source docs
Time on slides: 15 minutes, not 7 hours.
→ Write specs in Notion
→ Hit the Gamma API endpoint
→ Get the presentation in 3 minutes
→ Share live link that updates with source docs
Time on slides: 15 minutes, not 7 hours.
But if you treat slides like code, everything changes.
Before:
→ Write specs in Notion
→ Spend 6+ hours recreating in PowerPoint
→ Update slides manually when specs change
Before:
→ Write specs in Notion
→ Spend 6+ hours recreating in PowerPoint
→ Update slides manually when specs change
November 13, 2025 at 4:32 PM
But if you treat slides like code, everything changes.
Before:
→ Write specs in Notion
→ Spend 6+ hours recreating in PowerPoint
→ Update slides manually when specs change
Before:
→ Write specs in Notion
→ Spend 6+ hours recreating in PowerPoint
→ Update slides manually when specs change
Time to build a feature?
3 days.
Time to make the slides?
7 hours.
3 days.
Time to make the slides?
7 hours.
November 13, 2025 at 4:32 PM
Time to build a feature?
3 days.
Time to make the slides?
7 hours.
3 days.
Time to make the slides?
7 hours.
We automated deployments, testing, and code reviews.
But we're still manually dragging boxes in slides like it's 2005.
But we're still manually dragging boxes in slides like it's 2005.
November 13, 2025 at 4:32 PM
We automated deployments, testing, and code reviews.
But we're still manually dragging boxes in slides like it's 2005.
But we're still manually dragging boxes in slides like it's 2005.
If you enjoyed this thread:
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
Subscribe | Hungry Minds
Get smarter about Software and AI in 5 minutes. Save 50+ hours/week with deep dives, trends and tools hand-picked from 100+ sources. Join 50K+ engineers from big tech to startups for 1 free email every Monday.
www.hungryminds.dev
November 12, 2025 at 11:30 PM
If you enjoyed this thread:
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
Are you using Cursor?
November 12, 2025 at 11:30 PM
Are you using Cursor?
6. Technical Challenges:
↳ Network overhead: Retries due to server load spike traffic
↳ AST parsing edge cases: Language-specific syntax quirks
↳ Embedding inversion risks: Theoretical code leaks from vectors (mitigated by short TTLs)
↳ Network overhead: Retries due to server load spike traffic
↳ AST parsing edge cases: Language-specific syntax quirks
↳ Embedding inversion risks: Theoretical code leaks from vectors (mitigated by short TTLs)
November 12, 2025 at 11:30 PM
6. Technical Challenges:
↳ Network overhead: Retries due to server load spike traffic
↳ AST parsing edge cases: Language-specific syntax quirks
↳ Embedding inversion risks: Theoretical code leaks from vectors (mitigated by short TTLs)
↳ Network overhead: Retries due to server load spike traffic
↳ AST parsing edge cases: Language-specific syntax quirks
↳ Embedding inversion risks: Theoretical code leaks from vectors (mitigated by short TTLs)
5. Why Merkle trees?
↳ Bandwidth savings: Sync only delta changes (Git-like)
↳ Cache optimization: Hash-indexed embeddings enable instant re-indexing
↳ Data integrity: Tamper-proof codebase fingerprints
↳ Bandwidth savings: Sync only delta changes (Git-like)
↳ Cache optimization: Hash-indexed embeddings enable instant re-indexing
↳ Data integrity: Tamper-proof codebase fingerprints
November 12, 2025 at 11:30 PM
5. Why Merkle trees?
↳ Bandwidth savings: Sync only delta changes (Git-like)
↳ Cache optimization: Hash-indexed embeddings enable instant re-indexing
↳ Data integrity: Tamper-proof codebase fingerprints
↳ Bandwidth savings: Sync only delta changes (Git-like)
↳ Cache optimization: Hash-indexed embeddings enable instant re-indexing
↳ Data integrity: Tamper-proof codebase fingerprints
4. RAG for code generation:
↳ Query vector DB (Turbopuffer) for relevant chunks
↳ Inject top matches into LLM context
↳ Generate answers using GPT-4 + codebase context
↳ Query vector DB (Turbopuffer) for relevant chunks
↳ Inject top matches into LLM context
↳ Generate answers using GPT-4 + codebase context
November 12, 2025 at 11:30 PM
4. RAG for code generation:
↳ Query vector DB (Turbopuffer) for relevant chunks
↳ Inject top matches into LLM context
↳ Generate answers using GPT-4 + codebase context
↳ Query vector DB (Turbopuffer) for relevant chunks
↳ Inject top matches into LLM context
↳ Generate answers using GPT-4 + codebase context
3. Embedding and privacy:
↳ Uses OpenAI’s text-embedding-3-small or custom code-specific models
↳ Obfuscates file paths with client-side encryption (e.g., src/utils . py → a1b2/c3d4/e5f6)
↳ No raw code stored, embeddings purged after request
↳ Uses OpenAI’s text-embedding-3-small or custom code-specific models
↳ Obfuscates file paths with client-side encryption (e.g., src/utils . py → a1b2/c3d4/e5f6)
↳ No raw code stored, embeddings purged after request
November 12, 2025 at 11:30 PM
3. Embedding and privacy:
↳ Uses OpenAI’s text-embedding-3-small or custom code-specific models
↳ Obfuscates file paths with client-side encryption (e.g., src/utils . py → a1b2/c3d4/e5f6)
↳ No raw code stored, embeddings purged after request
↳ Uses OpenAI’s text-embedding-3-small or custom code-specific models
↳ Obfuscates file paths with client-side encryption (e.g., src/utils . py → a1b2/c3d4/e5f6)
↳ No raw code stored, embeddings purged after request
2. Merkle tree construction:
↳ Local hashing: Compute SHA-256 hashes for all code chunks
↳ Tree sync: Compare root hash with server to identify changed files
↳Incremental uploads: Only modified chunks get re-embedded
Result: 90% fewer uploads vs. full re-indexing.
↳ Local hashing: Compute SHA-256 hashes for all code chunks
↳ Tree sync: Compare root hash with server to identify changed files
↳Incremental uploads: Only modified chunks get re-embedded
Result: 90% fewer uploads vs. full re-indexing.
November 12, 2025 at 11:30 PM
2. Merkle tree construction:
↳ Local hashing: Compute SHA-256 hashes for all code chunks
↳ Tree sync: Compare root hash with server to identify changed files
↳Incremental uploads: Only modified chunks get re-embedded
Result: 90% fewer uploads vs. full re-indexing.
↳ Local hashing: Compute SHA-256 hashes for all code chunks
↳ Tree sync: Compare root hash with server to identify changed files
↳Incremental uploads: Only modified chunks get re-embedded
Result: 90% fewer uploads vs. full re-indexing.
1. Code chunking strategies:
↳ Code splitting: AST tree-sitter to parse code into logical blocks (functions, classes)
↳ Token limits: Merge sibling AST nodes without exceeding model token capacity
↳ Semantic boundaries: Avoid mid-function splits for better embeddings
↳ Code splitting: AST tree-sitter to parse code into logical blocks (functions, classes)
↳ Token limits: Merge sibling AST nodes without exceeding model token capacity
↳ Semantic boundaries: Avoid mid-function splits for better embeddings
November 12, 2025 at 11:30 PM
1. Code chunking strategies:
↳ Code splitting: AST tree-sitter to parse code into logical blocks (functions, classes)
↳ Token limits: Merge sibling AST nodes without exceeding model token capacity
↳ Semantic boundaries: Avoid mid-function splits for better embeddings
↳ Code splitting: AST tree-sitter to parse code into logical blocks (functions, classes)
↳ Token limits: Merge sibling AST nodes without exceeding model token capacity
↳ Semantic boundaries: Avoid mid-function splits for better embeddings
0. Merkle trees 101:
↳ Hierarchical hash chains that fingerprint data blocks
↳ Leaf nodes = hash of code chunks
↳ Parent nodes = hash of child hashes
↳ Root hash = single fingerprint for the entire codebase
Key benefit: Detect changes instantly by comparing root hashes.
↳ Hierarchical hash chains that fingerprint data blocks
↳ Leaf nodes = hash of code chunks
↳ Parent nodes = hash of child hashes
↳ Root hash = single fingerprint for the entire codebase
Key benefit: Detect changes instantly by comparing root hashes.
November 12, 2025 at 11:30 PM
0. Merkle trees 101:
↳ Hierarchical hash chains that fingerprint data blocks
↳ Leaf nodes = hash of code chunks
↳ Parent nodes = hash of child hashes
↳ Root hash = single fingerprint for the entire codebase
Key benefit: Detect changes instantly by comparing root hashes.
↳ Hierarchical hash chains that fingerprint data blocks
↳ Leaf nodes = hash of code chunks
↳ Parent nodes = hash of child hashes
↳ Root hash = single fingerprint for the entire codebase
Key benefit: Detect changes instantly by comparing root hashes.
Cursor uses Merkle trees as a core data structure.
November 12, 2025 at 11:30 PM
Cursor uses Merkle trees as a core data structure.
If you enjoyed this thread:
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
Subscribe to Hungry Minds
Get smarter about Software and AI in 5 minutes.
Save 50+ hours/week with deep dives, trends, and tools hand-picked from 100+ sources.
Join 50,000+ engineers from big tech to startups for 1 free email every Monday. Click to read Hungry Minds, by Alexandre Zajac, a Substack publication.
hungrymindsdev.substack.com
November 11, 2025 at 4:32 PM
If you enjoyed this thread:
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
1. Follow me @itsalexzajac for more content on Software and AI
2. Check out Hungry Minds: newsletter.hungryminds.dev
3. RT the tweet below to share this thread with your audience
What practices makes code easier to read?
November 11, 2025 at 4:32 PM
What practices makes code easier to read?
7. Avoid gotos:
↳ Use gotos only when absolutely necessary.
✅ They can make your code harder to follow and debug.
↳ Use gotos only when absolutely necessary.
✅ They can make your code harder to follow and debug.
November 11, 2025 at 4:32 PM
7. Avoid gotos:
↳ Use gotos only when absolutely necessary.
✅ They can make your code harder to follow and debug.
↳ Use gotos only when absolutely necessary.
✅ They can make your code harder to follow and debug.
6. Write smaller functions:
↳ Keep your functions small and focused on a single task.
✅ Smaller functions are easier to read, test, and reuse.
↳ Keep your functions small and focused on a single task.
✅ Smaller functions are easier to read, test, and reuse.
November 11, 2025 at 4:32 PM
6. Write smaller functions:
↳ Keep your functions small and focused on a single task.
✅ Smaller functions are easier to read, test, and reuse.
↳ Keep your functions small and focused on a single task.
✅ Smaller functions are easier to read, test, and reuse.
5. Keep variables short-lived:
↳ Declare variables as close to their usage as possible.
✅ This reduces the mental effort needed to track their values.
↳ Declare variables as close to their usage as possible.
✅ This reduces the mental effort needed to track their values.
November 11, 2025 at 4:32 PM
5. Keep variables short-lived:
↳ Declare variables as close to their usage as possible.
✅ This reduces the mental effort needed to track their values.
↳ Declare variables as close to their usage as possible.
✅ This reduces the mental effort needed to track their values.
4. Use distinct variable names:
↳ Choose descriptive and unique names for your variables.
✅ Avoid names that look similar, like `i` and `j`, to prevent confusion.
↳ Choose descriptive and unique names for your variables.
✅ Avoid names that look similar, like `i` and `j`, to prevent confusion.
November 11, 2025 at 4:32 PM
4. Use distinct variable names:
↳ Choose descriptive and unique names for your variables.
✅ Avoid names that look similar, like `i` and `j`, to prevent confusion.
↳ Choose descriptive and unique names for your variables.
✅ Avoid names that look similar, like `i` and `j`, to prevent confusion.