Links

Links for September 18, 2026

1. "Filtering with Inverted Indexes" "sql: implement zig-zag merge join #23520"
2. "SPANN: Highly-efficient Billion-scale Approximate Nearest Neighbor Search" "SPFresh: Incremental In-Place Update for Billion-Scale Vector Search" "RaBitQ: Quantizing High-Dimensional Vectors with a Theoretical Error Bound for Approximate Nearest Neighbor Search" "BtrBlocks: Efficient Columnar Compression for Data Lakes"

   Shared in the Turbopuffer tech talk

3. "Gowers's Weblog" "Why I didn’t sign the Fields medallists’ letter | Hacker News"
4. "Toward Recursive Self-Improvement: How GLM Built Its Own Inference Infrastructure" "How GLM built its own inference infrastructure | Hacker News"

   "We call this approach "dense feedback."...

   First, feedback must be sufficiently local. Wherever possible, it should be tied to specific engine launch parameters, code changes, kernels, input conditions, threads, execution intervals, or code paths, helping the agent narrow the scope of the problem. For example, instead of reporting that "model accuracy dropped after a fusion optimization," identifying the output differences for a specific request before and after the change gives the agent a better basis for constructing a minimal reproduction and analyzing the cause. ⁠Second, feedback must be inexpensive and timely to obtain. Whenever the agent proposes a hypothesis, makes a change, or constructs a set of controlled experiments, it should have an appropriate way to validate it. Questions that can be answered by a kernel test or a local microbenchmark should not require a full service deployment and end-to-end load test every time. Shorter validation cycles help the agent correct course promptly and spend less effort on unproductive hypotheses. ⁠Third, feedback must support objective verification. Whether a change is correct and whether performance has improved should be determined by reference implementations, test results, and comparable experimental metrics. Runtime signals can help the agent identify possible causes, but correlations between observations alone cannot establish a root cause. Controlled experiments are still needed to verify whether a change to a specific path produces the expected effect.”
5. "Hacking OpenAI" "A heap overflow and SSO misconfiguration to compromise OpenAI internal repos | Hacker News"
6. "The Best Philosophy Article of 2025 is Bad Economics"
7. "Education as an AI safety area"
8. "There's no point at which turning your brain off will work"

   Pretty related!

   "why aren't you pushing agents well beyond tasks they can tackle so easily?"

   Agreed! Though they need affordances (verification, observability - see GLM article above) so focus on those. Someone else phrased it as "being the platform team for your AI"