Nick Lothian
@nlothian.bsky.social
Gaslighting from Opus. Grr
February 2, 2026 at 3:52 AM
Gaslighting from Opus. Grr
Kimi K2.5 looks very impressive. 100 agent co-ordination, 4.5x speedup on tasks.
It's only January and we have a half order of magnitude improvement in real-world performance, and yet people think progress is slowing down?
Really, we are running out of benchmarks.
It's only January and we have a half order of magnitude improvement in real-world performance, and yet people think progress is slowing down?
Really, we are running out of benchmarks.
January 27, 2026 at 7:29 AM
Kimi K2.5 looks very impressive. 100 agent co-ordination, 4.5x speedup on tasks.
It's only January and we have a half order of magnitude improvement in real-world performance, and yet people think progress is slowing down?
Really, we are running out of benchmarks.
It's only January and we have a half order of magnitude improvement in real-world performance, and yet people think progress is slowing down?
Really, we are running out of benchmarks.
This could be a problem for random Australian politicians.
For those who don't know, in 2017-18 7 Australian politicians had to resign and/or run again because it turned out they could be dual citizens (sometimes unknowingly).
This change in Canadian law could trigger more!
For those who don't know, in 2017-18 7 Australian politicians had to resign and/or run again because it turned out they could be dual citizens (sometimes unknowingly).
This change in Canadian law could trigger more!
Scour your family tree for any hint of Canadian ancestors. Recently enacted Bill C-3 repealed the 1st gen limit on birth abroad & while most reporting implies a second gen limit it looks to really be more about how far back you can properly document the lineage.
January 20, 2026 at 11:43 AM
This could be a problem for random Australian politicians.
For those who don't know, in 2017-18 7 Australian politicians had to resign and/or run again because it turned out they could be dual citizens (sometimes unknowingly).
This change in Canadian law could trigger more!
For those who don't know, in 2017-18 7 Australian politicians had to resign and/or run again because it turned out they could be dual citizens (sometimes unknowingly).
This change in Canadian law could trigger more!
Tired: Sprint point estimates
Wired: Number of tokens to implement feature estimates
Wired: Number of tokens to implement feature estimates
January 15, 2026 at 4:27 AM
Tired: Sprint point estimates
Wired: Number of tokens to implement feature estimates
Wired: Number of tokens to implement feature estimates
Looking for feedback on a new HTML tool I created, heavily inspired by @simonwillison.net's fantastic collection of HTML tools.
LLM Comparator lets you try the same prompt on multiple LLMs, and continue the conversation across them.
tools.nicklothian.com/llm_comparat...
LLM Comparator lets you try the same prompt on multiple LLMs, and continue the conversation across them.
tools.nicklothian.com/llm_comparat...
January 6, 2026 at 6:10 AM
Looking for feedback on a new HTML tool I created, heavily inspired by @simonwillison.net's fantastic collection of HTML tools.
LLM Comparator lets you try the same prompt on multiple LLMs, and continue the conversation across them.
tools.nicklothian.com/llm_comparat...
LLM Comparator lets you try the same prompt on multiple LLMs, and continue the conversation across them.
tools.nicklothian.com/llm_comparat...
It shocks me how many people think LLM inference providers are somehow running at a loss.
Do the math! If you can hire a cloud H100 and run it above 50% capacity for something approaching the per token price inference providers are charging then it's easy to see they aren't.
Do the math! If you can hire a cloud H100 and run it above 50% capacity for something approaching the per token price inference providers are charging then it's easy to see they aren't.
December 24, 2025 at 6:35 AM
It shocks me how many people think LLM inference providers are somehow running at a loss.
Do the math! If you can hire a cloud H100 and run it above 50% capacity for something approaching the per token price inference providers are charging then it's easy to see they aren't.
Do the math! If you can hire a cloud H100 and run it above 50% capacity for something approaching the per token price inference providers are charging then it's easy to see they aren't.
OpenAI drama making crypto bros look like the sensible ones.
November 21, 2023 at 12:48 AM
OpenAI drama making crypto bros look like the sensible ones.
Reposted by Nick Lothian
🚀Happy to announce my recent open source project- LightPHE: A Lightweight Partially Homomorphic Encryption Library for #Python. It allows computations on encrypted data without decryption. It wraps many algorithms such as RSA, ElGamal, Paillier and more!
github.com/serengil/Lig...
github.com/serengil/Lig...
November 19, 2023 at 5:31 PM
🚀Happy to announce my recent open source project- LightPHE: A Lightweight Partially Homomorphic Encryption Library for #Python. It allows computations on encrypted data without decryption. It wraps many algorithms such as RSA, ElGamal, Paillier and more!
github.com/serengil/Lig...
github.com/serengil/Lig...
Reposted by Nick Lothian
We evaluate the combined architecture on two challenging planning tasks:
-Graph traversal
-Tower of Hanoi
Graph traversal:
significant improvements over zero-shot prompting or in-context learning.
Notably the architecture lowers hallucination of invalid moves to 0% (from ~20% for 4-step paths!)
4/n
-Graph traversal
-Tower of Hanoi
Graph traversal:
significant improvements over zero-shot prompting or in-context learning.
Notably the architecture lowers hallucination of invalid moves to 0% (from ~20% for 4-step paths!)
4/n
October 27, 2023 at 3:02 PM
We evaluate the combined architecture on two challenging planning tasks:
-Graph traversal
-Tower of Hanoi
Graph traversal:
significant improvements over zero-shot prompting or in-context learning.
Notably the architecture lowers hallucination of invalid moves to 0% (from ~20% for 4-step paths!)
4/n
-Graph traversal
-Tower of Hanoi
Graph traversal:
significant improvements over zero-shot prompting or in-context learning.
Notably the architecture lowers hallucination of invalid moves to 0% (from ~20% for 4-step paths!)
4/n
This seems like an appropriate first post.
Some high quality research into the frequency of compound pejoratives used on Reddit
colinmorris.github.io/blog/compoun...
colinmorris.github.io/blog/compoun...
October 19, 2023 at 1:52 AM
This seems like an appropriate first post.