toughcoding

REAL HUMAN

[toughcoding @toughcoding]

Secure your Bunny net streams & downloads—17% off! Use code TOUGHCODINGYT17
at 👉 https://t.co/Xu8C8pjtSX • 🔐 Signed CDN URLs to lock down file downloads
• 🎥 Token‑protected video embeds—no hotlinking, no leaks

No fluff. Just airtight content protection. ⚡️

[toughcoding @toughcoding]

https://t.co/hlOTcuu3YK

[toughcoding @toughcoding]

I am glad people do not use AI to generate fluff in @LinkedIn section comments - you can spot difference in language between post and people comments and even author post and his own comment. The fluffy posts are like marketing bs - maximalism in text but minimalism in meaning.

[toughcoding @toughcoding]

🚨 Linux Privilege Escalation Alert 🚨

“Copy Fail” (CVE-2026-31431) exposes a serious flaw in the Linux kernel, allowing local users to gain root via page cache writes. No network access or debugging needed—just a low-privileged account.

PoC: https://t.co/OLGLXh9wkK
Patch ASAP.

[toughcoding @toughcoding]

Graph ML goes beyond RAG by embedding structure (n-hop context), not just text. Agentic systems use KGs but don’t learn on them yet—bridging graph ML and agents is a key opportunity.

[toughcoding @toughcoding]

https://t.co/LyCk60KD1D
@mattpocockuk
Keeping large components with simple interfaces raising a question who is implementing this components. if llm cannot make it right then I will have placeholder with simple mocked interface but not real solution, tdd will be like assert true

[toughcoding @toughcoding]

@everestchris6 instead typing here in comment section I typed in claude code team and ask for same giving screenshot as reference. then I got complete system in the morning

[toughcoding @toughcoding]

@everythingLLM ollama --verbose run gemma-4-26b-a4b-it-mxfp4:latest
⠹ Error: 500 Internal Server Error: mlx runner failed: Error: unsupported architecture: Gemma4ForConditionalGeneration (exit: exit status 1)

I already noticed it.

[toughcoding @toughcoding]

They said Ollama support mlx ... so I tried to pickup some model that I used with mlx_lm and it failed.

#ollama #mlx

[toughcoding @toughcoding]

Testing Docker MCP Toolkit.
Is it something for MCP Clients like LiteLLM for LLM Clients. Although intentionally you are going to run that 'LLM'

[toughcoding @toughcoding]

Need to check it.

[toughcoding @toughcoding]

#aws #bahrain

[toughcoding @toughcoding]

@elastic_devs

Sadly speaking, why is not there yet ...

[toughcoding @toughcoding]

@elastic_devs
Serverless deployment seems cutting off automatically access to LLMs for unknown reasons.
0.620603 1M Tokens usage is far from being huge.

[toughcoding @toughcoding]

if anyone wonder hyped LuxTTS is useless.

I staying with F5-TTS - not fine tuned is great.

fine tuned F5-TTS is even better as removing chinese accent

[toughcoding @toughcoding]

🇺🇦☢️ 189 radiation stations. 3 messy JS files. 1 @elastic AI agent. Parses, merges, indexes, analyzes, and publishes a daily safety report — zero manual steps. Built for the #ElasticHackathon with Workflows + Agent Builder.
https://t.co/cCcx9TWwQh

[toughcoding @toughcoding]

what you told me yesterday I am thinking how avoid self-affirming hallucination machine when building agentic systems.

[toughcoding @toughcoding]

@elonmusk Clawdbot or ClownBot ?

[toughcoding @toughcoding]

@DuckyCloud0 @ujjwalscript indeed AI learn from humans so

[toughcoding @toughcoding]

@setupspawn indeed, just tried it

[toughcoding @toughcoding]

@PhookyCom @AlexFinn Indeed. Most people do not disclose how much tokens they are burning - that's for sure less than hiring human being but certainly not cheap.