close

DEV Community

Maxim Saplin profile picture

Maxim Saplin

ツ AI in Software Dev, Open-source

Education

Computer Science @BNTU, MBA @BSU

Work

EPAM, Delivery Partner

Top 7
6
Google AI
Six Year Club
Five Year Club
Writing Debut
16 Week Writing Streak
Four Year Club
8 Week Writing Streak
4 Week Community Wellness Streak
2 Week Community Wellness Streak
4 Week Writing Streak
1 Week Community Wellness Streak
The AI Bill Grows in the Agent Loop

Retries and loops as primary cost drivers

The AI Bill Grows in the Agent Loop

BERJAYA BERJAYA BERJAYA 16
Comments 11
16 min read

Want to connect with Maxim Saplin?

Create an account to connect with Maxim Saplin. You can also sign in below to proceed if you already have an account.

Already have an account? Sign in
CLI over MCP: a small Chrome DevTools experiment in Copilot CLI

Avoids 5k tokens of upfront MCP schema bloat

CLI over MCP: a small Chrome DevTools experiment in Copilot CLI

BERJAYA BERJAYA BERJAYA 14
Comments 9
8 min read
Debloating The AI-Grown Codebase

31.7% reduction with tests still green

Debloating The AI-Grown Codebase

BERJAYA BERJAYA BERJAYA 33
Comments 18
9 min read
AI Agent Failure Modes Beyond Hallucination

Taxonomy of amnesia and recursive cost drift

AI Agent Failure Modes Beyond Hallucination

BERJAYA BERJAYA BERJAYA 35
Comments 28
7 min read
AI Agents vs Code Vulnerabilities: Was Anthropic Mythos a Big Deal or Fear-mongering?

AI Agents vs Code Vulnerabilities: Was Anthropic Mythos a Big Deal or Fear-mongering?

BERJAYA BERJAYA BERJAYA 19
Comments 8
12 min read
Long-Horizon Agents Are Here. Full Autopilot Isn't

Small tasks exposing fragile model loops

Long-Horizon Agents Are Here. Full Autopilot Isn't

BERJAYA BERJAYA BERJAYA 33
Comments 18
7 min read
Ran out of Cursor tokens and switched to GitHub Copilot: Side-by-Side

Ran out of Cursor tokens and switched to GitHub Copilot: Side-by-Side

BERJAYA BERJAYA BERJAYA 29
Comments 22
9 min read
Long-horizon agents: OpenCode + GPT-5.2 Codex Experiment

Long-horizon agents: OpenCode + GPT-5.2 Codex Experiment

BERJAYA BERJAYA BERJAYA 9
Comments
3 min read
Cursor-like Semantic Rules in GitHub Copilot

Cursor-like Semantic Rules in GitHub Copilot

BERJAYA BERJAYA BERJAYA 8
Comments
2 min read
AI Dev: Plan Mode vs. SDD — A Weekend Experiment

AI Dev: Plan Mode vs. SDD — A Weekend Experiment

BERJAYA BERJAYA BERJAYA 16
Comments 4
8 min read
AI Dev: Testing Kiro

AI Dev: Testing Kiro

BERJAYA BERJAYA BERJAYA 19
Comments 10
7 min read
LLMs are Bad at Math

LLMs are Bad at Math

BERJAYA BERJAYA BERJAYA 14
Comments 1
6 min read
Grok 3 API - Reasoning Tokens are Counted Differently

Grok 3 API - Reasoning Tokens are Counted Differently

BERJAYA BERJAYA BERJAYA 10
Comments
1 min read
XYZ% of Code is Now Written by AI... Who Cares?

XYZ% of Code is Now Written by AI... Who Cares?

BERJAYA BERJAYA BERJAYA 43
Comments 8
6 min read
GPT 4.1, o3, o4-mini - OpenAI releases through the lens of LLM_Chess

GPT 4.1, o3, o4-mini - OpenAI releases through the lens of LLM_Chess

BERJAYA BERJAYA BERJAYA 10
Comments
1 min read
Mercury Coder - A Quick Test of Diffusion Language Model

Mercury Coder - A Quick Test of Diffusion Language Model

BERJAYA BERJAYA BERJAYA 10
Comments
4 min read
Llama 4 - 10M Context? Coding? Decent Follow-up?

Llama 4 - 10M Context? Coding? Decent Follow-up?

BERJAYA BERJAYA BERJAYA 25
Comments 1
6 min read
4o Image Gen - Diffusion/Transformer Cross-over Trend?

4o Image Gen - Diffusion/Transformer Cross-over Trend?

BERJAYA BERJAYA BERJAYA 20
Comments
3 min read
SGLang vs llama.cpp - A Quick Speed Test

SGLang vs llama.cpp - A Quick Speed Test

BERJAYA BERJAYA BERJAYA 20
Comments 3
3 min read
OpenAI o3-mini Tested in LLM Chess

OpenAI o3-mini Tested in LLM Chess

BERJAYA BERJAYA BERJAYA 16
Comments
5 min read
Qwen2.5 Max Release Went Unnoticed in Deepseek Hysteria

Qwen2.5 Max Release Went Unnoticed in Deepseek Hysteria

BERJAYA BERJAYA BERJAYA 11
Comments
6 min read
DeepSeek-V3: Laziness vs Eagerness

DeepSeek-V3: Laziness vs Eagerness

BERJAYA BERJAYA BERJAYA 19
Comments
2 min read
OpenAI o1/o3 - Be Careful What you Wish For...

OpenAI o1/o3 - Be Careful What you Wish For...

BERJAYA BERJAYA BERJAYA 9
Comments 2
2 min read
Tried Phi-4, It didn't Impress

Tried Phi-4, It didn't Impress

BERJAYA BERJAYA 23
Comments
2 min read
Gemini 2.0 Released, Reminding of "AI Hitting the Wall" Talks

Gemini 2.0 Released, Reminding of "AI Hitting the Wall" Talks

BERJAYA BERJAYA BERJAYA 13
Comments 1
2 min read
The Power of Pragmatism: Engineering Cultures and China's Ascendancy

The Power of Pragmatism: Engineering Cultures and China's Ascendancy

BERJAYA BERJAYA BERJAYA 8
Comments
4 min read
Can LLMs Play Chess? I've Tested 13 Models (GPT-4o, Claude 3.5, Gemini 1.5 etc.)

Can LLMs Play Chess? I've Tested 13 Models (GPT-4o, Claude 3.5, Gemini 1.5 etc.)

BERJAYA BERJAYA BERJAYA 44
Comments 9
8 min read
Microsoft Autogen Has Split in 2... Wait 3... No, 4 Parts

Microsoft Autogen Has Split in 2... Wait 3... No, 4 Parts

BERJAYA BERJAYA BERJAYA 44
Comments 12
3 min read
Llama 3.1 Nemotron 70B - Quirks and Features

Llama 3.1 Nemotron 70B - Quirks and Features

BERJAYA BERJAYA BERJAYA 54
Comments 2
7 min read
DDR5 Speed, CPU and LLM Inference

DDR5 Speed, CPU and LLM Inference

BERJAYA BERJAYA BERJAYA 26
Comments 5
4 min read
Gen AI Hype - the Never Ending Excitement

Gen AI Hype - the Never Ending Excitement

BERJAYA BERJAYA BERJAYA 17
Comments 1
8 min read
OpenAI o1 Release is so Reminiscent of Apple Events - it's an Incremental Update

OpenAI o1 Release is so Reminiscent of Apple Events - it's an Incremental Update

BERJAYA BERJAYA BERJAYA 14
Comments 4
7 min read
Continue.dev: The Swiss Army Knife That Sometimes Fails to Cut

Continue.dev: The Swiss Army Knife That Sometimes Fails to Cut

BERJAYA BERJAYA BERJAYA 61
Comments 4
8 min read
Python 3.13 RC1 - a Quick CPU Benchmark

Python 3.13 RC1 - a Quick CPU Benchmark

BERJAYA BERJAYA 11
Comments 1
2 min read
llama.cpp: CPU vs GPU, shared VRAM and Inference Speed

llama.cpp: CPU vs GPU, shared VRAM and Inference Speed

BERJAYA BERJAYA BERJAYA 38
Comments 7
3 min read
Convergence of LLMs: 2024 Trend Solidified by Llama 3.1 Release

Convergence of LLMs: 2024 Trend Solidified by Llama 3.1 Release

BERJAYA BERJAYA BERJAYA 14
Comments 4
3 min read
DoLa and MT-Bench - A Quick Eval of a new LLM trick

DoLa and MT-Bench - A Quick Eval of a new LLM trick

BERJAYA BERJAYA 6
Comments
2 min read
4090 - ECC ON vs ECC OFF

4090 - ECC ON vs ECC OFF

BERJAYA BERJAYA 11
Comments
1 min read
MT-Bench: Comparing different LLM Judges

MT-Bench: Comparing different LLM Judges

BERJAYA BERJAYA BERJAYA 19
Comments 2
4 min read
Nvidia's 1000x Performance Boost Claim Verified

Nvidia's 1000x Performance Boost Claim Verified

BERJAYA BERJAYA BERJAYA 12
Comments
2 min read
LLM Fine-tunig on RTX 4090: 90% Performance at 55% Power

LLM Fine-tunig on RTX 4090: 90% Performance at 55% Power

BERJAYA BERJAYA BERJAYA 28
Comments
2 min read
GPT-4o: sneak peak at Llama 3 400B and the Age of Loneliness...

GPT-4o: sneak peak at Llama 3 400B and the Age of Loneliness...

BERJAYA BERJAYA BERJAYA 24
Comments 1
3 min read
FineWeb 45TB Dataset: $500k GPU costs and Adult Content Improving LLM Quality

FineWeb 45TB Dataset: $500k GPU costs and Adult Content Improving LLM Quality

BERJAYA BERJAYA BERJAYA 14
Comments 1
2 min read
Llama 3 8B is better than Llama 2 70B

Llama 3 8B is better than Llama 2 70B

BERJAYA BERJAYA BERJAYA 25
Comments
1 min read
Fine-tuning LLM on a laptop: VRAM - Shared Memory - GPU Load - Performance

Fine-tuning LLM on a laptop: VRAM - Shared Memory - GPU Load - Performance

BERJAYA BERJAYA BERJAYA 10
Comments
3 min read
3 Manifestations of GenAI in Software Development

3 Manifestations of GenAI in Software Development

BERJAYA BERJAYA 6
Comments
1 min read
AI-assisted coding, Sleeping on a Volcano

AI-assisted coding, Sleeping on a Volcano

BERJAYA BERJAYA BERJAYA 19
Comments 1
9 min read
LLM's "commendable, innovative, meticulous, notable, versatile, intricate" impact

LLM's "commendable, innovative, meticulous, notable, versatile, intricate" impact

BERJAYA BERJAYA BERJAYA 14
Comments
2 min read
Running Local LLMs, CPU vs. GPU - a Quick Speed Test

Running Local LLMs, CPU vs. GPU - a Quick Speed Test

BERJAYA BERJAYA BERJAYA 351
Comments 49
4 min read
Apple is killing PWA?

Apple is killing PWA?

BERJAYA BERJAYA BERJAYA 42
Comments 15
8 min read
Memories in ChatGPT: Privacy Implications

Memories in ChatGPT: Privacy Implications

BERJAYA BERJAYA BERJAYA 8
Comments
3 min read
Apple Vision Pro is the best marketing campaign for Meta's Quest 3

Apple Vision Pro is the best marketing campaign for Meta's Quest 3

BERJAYA BERJAYA BERJAYA 12
Comments 4
2 min read
⟨ Cursor.sh ⟩ - a competitor to GitHub CoPilot

⟨ Cursor.sh ⟩ - a competitor to GitHub CoPilot

BERJAYA BERJAYA BERJAYA 152
Comments
10 min read
Google's Slow Burn: Project IDX's Half-Year Echo

Google's Slow Burn: Project IDX's Half-Year Echo

BERJAYA BERJAYA BERJAYA 11
Comments
2 min read
C#, Dart, TypeScript , Python: side-by-side

C#, Dart, TypeScript , Python: side-by-side

BERJAYA BERJAYA BERJAYA 31
Comments 10
7 min read
Chatting with AI: The Freedom of Private Interfaces

Chatting with AI: The Freedom of Private Interfaces

BERJAYA BERJAYA BERJAYA 15
Comments 2
8 min read
Phi-2 is available for chat through LM Studio Beta

Phi-2 is available for chat through LM Studio Beta

BERJAYA BERJAYA BERJAYA 10
Comments
2 min read
How fast is JS tiktoken?

How fast is JS tiktoken?

BERJAYA BERJAYA BERJAYA 10
Comments
1 min read
`Get Abstract` for Lex Fridman and Jeff Bezos talk (December 2023)

`Get Abstract` for Lex Fridman and Jeff Bezos talk (December 2023)

BERJAYA BERJAYA BERJAYA 8
Comments
3 min read
Python 3.12 Performance - a Quick Test

Python 3.12 Performance - a Quick Test

BERJAYA BERJAYA BERJAYA 44
Comments 6
2 min read
loading...