Logesh @logesh_io

software engineer. curious about how things work, why they break, and how to make them scale. Coimbatore Joined May 2024

Tweets

20
Followers

0
Following

34
Likes

4

Logesh @logesh_io

15 hours ago

@mrkaran_ Nice direction. Now i am more curious how Logchef is going to route queries across hot and cold storage.

0 0 0 10 0

View Details

The pace at which these models are improving is getting hard to ignore. Open models are no longer just catching up. The are starting to change what teams should build themselves vs pay an API for. Now let’s see how OpenAI and Anthropic respond.

Kimi.ai @Kimi_Moonshot

18 hours ago

Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model. 🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal. 🔹 Full-project coding: games in Unreal, Unity & Godot. 🔹 Excels at parametric 3D modeling in Blender & CAD. 🔹 More reliable at

1K 3K 26K 2.6M 5K

0 0 0 9 0

View Details

Logesh @logesh_io

17 hours ago

Useful debugging reminder: df -h → disk/block usage df -i → inode usage “Disk full” is not always about bytes. Sometimes the filesystem has space, but nowhere left to create another file.

0 0 0 6 0

View Details

Logesh @logesh_io

17 hours ago

This is easy to hit with workloads that create a lot of small files. node_modules is a good example. You can run out of inodes long before you run out of storage capacity.

1 0 0 7 0

View Details

Logesh @logesh_io

17 hours ago

npm install failed on one of our servers with: ENOSPC: no space left on device First check was obvious: df -h Plenty of disk space. So the error looked wrong.

1 0 0 13 0

View Details

Logesh @logesh_io

22 hours ago

@OliverBugar Play it safe, peace over hustle.

0 0 1 6 0

View Details

Logesh @logesh_io

23 hours ago

@Ayush_builds_ B. Great product can evolve over the time

0 0 0 9 0

View Details

Logesh @logesh_io

23 hours ago

@mmalisper @pgrustdb 3x faster than ClickHouse caught my attention. Is there a benchmark repo or detailed comparison available somewhere?

0 0 0 22 0

View Details

Logesh @logesh_io

a day ago

@thomasunise And now AI is going to help us generate that same slop 10x faster with even fewer people understanding what the code actually does.

0 0 0 4 0

View Details

Logesh @logesh_io

a day ago

Every index is another data structure the database must maintain. Reads may get faster because there's a shorter path to the rows. But inserts, deletes, and some updates now have extra work to do. Indexes should follow query patterns, not column count.

0 0 0 11 0

View Details

Logesh @logesh_io

a day ago

@VadimStrizheus couldn't agree more

0 0 0 21 0

View Details

Logesh @logesh_io

a day ago

@benswerd Winning developer mindshare and winning production workloads are two very different things.

0 0 0 18 0

View Details

Logesh @logesh_io

2 days ago

@livingdevops Hi

0 0 0 4 0

View Details

Logesh @logesh_io

2 days ago

@piyushgarg_dev ChatGPT somehow feels less judgmental 😂

0 0 0 116 0

View Details

Logesh @logesh_io

2 days ago

@alxshp Per-agent databases look promising for isolation. How are you handling lifecycle management and cleanup when hundreds of agents spin them up at once?

0 0 0 3 0

View Details

Logesh @logesh_io

2 days ago

@livingdevops I think workload unpredictability matters more than traffic volume. Short-lived, bursty agent runs complicate scheduling, node churn and efficient scale-down. Kubernetes makes sense when you need its scheduling capabilities; app growth alone doesn’t justify it.

0 0 0 2 0

View Details

Logesh @logesh_io

2 days ago

@thdxr The hard part isn’t figuring out what each tool does. It’s knowing who owns it, what relies on it, and what happens when it breaks. Agents can help map things out, but ownership and accountability still need engineers.