Same here, I type way more now reviewing and prompting than I ever did just coding. The frustration is much higher since you do not fully control the output compared to when you just copy-pasted from a quick google search.
@whats_ai
-
Late Interaction Architecture for Large-Scale Document Retrieval Systems
By
–
Late interaction seems like it could be the sweet spot between full cross-attention and bi-encoders. Curious how it scales with really large document collections in production, and if we really want that in any case vs. optimizing compaction methods 🙂
-
Figma MCP Agents Transform Design-to-Code Workflows
By
–
Figma + MCP is a great combo. Being able to have agents interact with designs directly instead of just screenshots is next level for design-to-code workflows.
-
Claude outperforms GPT on coding and long context tasks
By
–
On my end, Claude's models usually hold up quite nicely vs. GPT's. Especially for coding and long context tasks. The growth makes sense.
-
AI agents management: delegating control over code generation
By
–
The shift from "AI writes my code" to "AI manages the agents that write my code" is happening faster than I expected. The meta-skill is knowing what to delegate vs what to control.
-
Supply Chain Attacks Threaten AI Development Pipelines
By
–
Supply chain attacks on AI tooling is a bit alarming TBH. Most companies just pip install and deploy without auditing dependencies… scary failure mode.
-
Moxie Brings Encrypted AI to Meta Platform
By
–
Moxie bringing encrypted AI to Meta is genuinely one of the biggest under-the-radar announcements this year. The fact the algo buried it says everything about what gets rewarded vs what actually matters.
-
FlashAttention-4 Release Bridges Gap Between Research and Production
By
–
FlashAttention-4 in the official release is huge. The gap between research implementations and production-ready code keeps shrinking. Also nice to see MPS getting more love, Apple Silicon users have been waiting.
-
Claude Claws and Cowork Transform Daily Workflow Experience
By
–
Ok "Claude Claws" is genuinely funny. Also been using Cowork all day and it's already changing how I work, the computer use part is surprisingly smooth.
-
Human-in-the-Loop: The Core Feature of Safe AI Agents
By
–
This is exactly it. The approval loop is the entire product. Nobody cares how powerful your agent is if they're scared to let it run unsupervised. Human-in-the-loop isn't a limitation, it's the feature.