toughcoding @toughcoding

REAL HUMAN https://t.co/yZtYx2zoch https://t.co/xNupSaZmpT youTube: https://t.co/ti6eBwOcT0 toughcoding.net Joined April 2023

Tweets

282
Followers

12
Following

80
Likes

18

toughcoding @toughcoding

a month ago

anthropic.com/careers/jobs

0 0 0 25 0

View Details

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.

0 0 0 20 0

View Details

toughcoding @toughcoding

a month ago

🚨 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: github.com/theori-io/copy… Patch ASAP.

0 0 0 151 0

View Details

toughcoding @toughcoding

a month ago

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.

0 0 0 15 0

View Details

toughcoding @toughcoding

a month ago

youtu.be/v4F1gFy-hqg @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

0 0 0 23 0

View Details

toughcoding @toughcoding

2 months ago

@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

0 0 2 1K 0

View Details

toughcoding @toughcoding

2 months ago

@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.

0 0 0 71 0

View Details

toughcoding @toughcoding

2 months ago

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

1 0 0 67 0

View Details

toughcoding @toughcoding

2 months ago

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

0 0 0 33 0

View Details

toughcoding @toughcoding

2 months ago

Need to check it.

0 0 0 18 0

View Details

toughcoding @toughcoding

3 months ago

#aws #bahrain

0 0 0 114 0

View Details

toughcoding @toughcoding

3 months ago

@elastic_devs Sadly speaking, why is not there yet ...

0 0 0 3 0

View Details

toughcoding @toughcoding

3 months ago

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

0 0 0 2 0

View Details

toughcoding @toughcoding

3 months ago

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

0 0 0 33 0

View Details

toughcoding @toughcoding

3 months ago

🇺🇦☢️ 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. github.com/toughcoding/es…

0 0 1 64 0

View Details

toughcoding @toughcoding

4 months ago

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

0 0 0 36 0

View Details

toughcoding @toughcoding

4 months ago

@elonmusk Clawdbot or ClownBot ?

0 0 0 7 0

View Details

toughcoding @toughcoding

4 months ago

@DuckyCloud0 @ujjwalscript indeed AI learn from humans so

0 0 0 758 0

View Details

toughcoding @toughcoding

4 months ago

@setupspawn indeed, just tried it

1 1 3 3K 2

View Details

toughcoding @toughcoding

4 months ago

@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.