Lars Winstand - DEV Community

Lars Winstand

Oct 16

I finally get why people ask for an uncensored LLM API, and it’s not for the reason everyone assumes

#ai #llm #automation #api

7 min read

Lars Winstand

Oct 16

I gave my n8n agent email access and realized one bad PDF could make it go rogue

#ai #n8n #security #automation

1

7 min read

Lars Winstand

Oct 16

My WhatsApp bot not replying problem had almost nothing to do with GPT-5

#whatsapp #api #webdev #ai

8 min read

Lars Winstand

Oct 15

A Claude Max billing error is what finally made me stop running agents on chat subscriptions

#ai #devops #automation #api

7 min read

Lars Winstand

Oct 14

I finally cut AI context bloat by 80% and the fix was embarrassingly simple

#ai #llm #agents #devops

6 min read

Lars Winstand

Oct 11

I thought GPT-4o got 80% faster, but it was just prompt caching messing with my benchmark

#ai #llm #benchmarking #devops

7 min read

Lars Winstand

Oct 10

Our chatbot not replying in groups turned out to be a Slack plumbing bug, not an AI bug

#slack #python #ai #devops

1

7 min read

Lars Winstand

Oct 3

My support bot stopped forgetting everything when I changed AI memory write triggers

#ai #n8n #agents #devops

5 min read

Lars Winstand

Oct 3

TIL the best automations reduce context window by using cheap models for boring steps and saving Claude Opus for the hard part

#ai #automation #llm #devops

8 min read

Lars Winstand

Oct 2

I stopped letting one agent do everything and my tool calls got way less cursed

#ai #python #automation #devops

1

6 min read

Lars Winstand

Oct 1

We thought our GPT-5.4 agent got lazier in production — it was a 3-bug workflow teaching it to quit

#ai #llm #n8n #devops

7 min read

Lars Winstand

Sep 30

I thought my n8n LLM bug was OpenAI and it turned out to be 5 boring workflow failures

#n8n #ai #openai #devops

7 min read

Lars Winstand

Sep 30

I thought my cheap AI workflow was fine until I counted 3,000 tiny calls

#ai #automation #n8n #openai

1

7 min read

Lars Winstand

Sep 29

My headless AI server setup stopped dying when I got boring about queues and 30-second watchdogs

#ai #devops #automation #node

8 min read

Lars Winstand

Sep 28

My Google AI API 500 errors stopped being scary when I stopped retrying the whole workflow

#ai #devops #python #n8n

1

7 min read

Lars Winstand

Sep 28

My background AI processes worked great in the demo and then fell apart the second they had to wait

#ai #devops #workflow #automation

1

7 min read

Lars Winstand

Sep 28

My OAuth login worked. My Google Ads automation still died at 3 AM.

#oauth #googleads #meta #automation

7 min read

Lars Winstand

Sep 27

I ripped out my vector DB and my internal agent got better answers from markdown

#ai #llm #rag #devops

1

6 min read

Lars Winstand

Sep 27

My support bot kept forgetting the last ticket until I stopped treating Slack and email like separate worlds

#ai #n8n #automation #developers

1

7 min read

Lars Winstand

Sep 27

I finally figured out why my Mac mini kept killing agents overnight when the screen was off

#macos #automation #n8n #devops

7 min read

Lars Winstand

Sep 25

My agent got way better when I stopped feeding it garbage first

#ai #llm #n8n #automation

7 min read

Lars Winstand

Sep 24

My AI bill wasn’t exploding from tokens — it was exploding because the same failed step ran 3 times

#ai #devops #automation #api

8 min read

Lars Winstand

Sep 24

My Telegram bot spam fix was realizing moderation is an agent workflow, not 1 GPT prompt

#telegram #ai #python #devops

8 min read

Lars Winstand

Sep 23

I used Jaeger trace debugging on my agent and found out the LLM wasn’t the problem at all

#opentelemetry #jaeger #ai #devops

7 min read

Lars Winstand

Sep 22

I thought prompt caching was free money until my agent stack broke it on every single call

#ai #llm #openai #anthropic

1

2

7 min read

Lars Winstand

Sep 21

My AI agent failed obvious tasks, and 49% fewer retrieval misses changed how I debugged it

#ai #llm #agents #rag

1

6 min read

Lars Winstand

Sep 21

I finally figured out what people mean by the best uncensored LLM after watching 17 agent workflows fail

#ai #llm #api #devops

7 min read

Lars Winstand

Sep 21

My AI cron job looked cheap at 5 runs/day. At 96 runs/day it turned into 1,728 model calls

#ai #automation #devops #n8n

5 min read

Lars Winstand

Sep 20

My AI agent self-healing fix was embarrassingly simple once I stopped retrying the whole thing

#ai #agents #n8n #devops

6 min read

Lars Winstand

Sep 20

My refund agent looked smart until it lied about Shopify, Stripe, and Zendesk 3 different ways

#ai #automation #devops #api

7 min read

Lars Winstand

Sep 20

We rebuilt 47-node n8n flows until we admitted script-first AI workflows are just better

#ai #n8n #automation #devops

7 min read

Lars Winstand

Sep 19

I tested VM vs Docker security for agent workers because “just use Docker” stopped feeling like an answer

#docker #security #devops #ai

8 min read

Lars Winstand

Sep 19

We kept arguing about GPT vs Claude and missed the 2x long-context cost blowing up our agent runs

#ai #llm #agents #devops

7 min read

Lars Winstand

Sep 19

OAuth redirect URI mismatch ate more time than my agent logic — 5 setup changes that finally fixed Google, Slack, and HubSpot

#oauth #node #webdev #api

8 min read

Lars Winstand

Sep 17

The most expensive part of my agent stack wasn’t tokens — it was the invoice roulette

#ai #agents #n8n #openai

7 min read

Lars Winstand

Sep 17

I built a headless Mac mini AI server and by week two I was debugging sleep, plist files, and Redis

#ai #macos #n8n #devops

7 min read

Lars Winstand

Sep 17

I thought lead routing was a prompt problem until the AI assigned a California lead to nobody

#ai #automation #n8n #devops

6 min read

Lars Winstand

Sep 16

I cut my agent bill by shrinking the prompt, not the model

#ai #llm #agents #devops

7 min read

Lars Winstand

Sep 15

The best n8n fix I found this month was boring: lower your agent concurrency settings before touching the prompt

#n8n #ai #automation #devops

5 min read

Lars Winstand

Sep 15

I thought the DeepSeek price increase wouldn’t matter until my n8n workflow turned 1 task into 3 billable model runs

#ai #automation #api #n8n

7 min read

Lars Winstand

Sep 14

My supervisor-worker agent pattern fixed the dumbest bug in my automation stack at 1:13 AM

#ai #agents #automation #python

1

6 min read

Lars Winstand

Sep 12

I kept blaming my prompt for the Meta Ads MCP 401, and it turned out Meta was rejecting the token for 3 totally different reasons

#api #meta #oauth #ai

7 min read

Lars Winstand

Sep 12

I thought prompt tweaks would fix my slow agent, but the real win was chasing cold starts first

#ai #devops #n8n #serverless

7 min read

Lars Winstand

Sep 11

Our AI agent pipeline kept stalling on CAPTCHA-heavy scraping until we rebuilt it around queues and short browser sessions

#webscraping #playwright #ai #automation

8 min read

Lars Winstand

Sep 9

I went looking for a DeepSeek API alternative and learned a 500-request cap is where cheap endpoints start breaking

#ai #api #automation #devops

1

7 min read

Lars Winstand

Sep 8

We finally learned to switch LLM providers with 3 code changes instead of rewriting half the app

#ai #llm #python #api

7 min read

Lars Winstand

Sep 8

I stopped treating local models like drop-in GPT replacements after 4 output bugs broke real automations

#ai #llm #automation #devops

6 min read

Lars Winstand

Sep 7

My tool-calling agent looked amazing in demos and then double-charged people

#ai #llm #testing #devops

7 min read

Lars Winstand

Sep 5

I thought local Whisper transcription would make call notes basically free, then a 13-minute file turned cleanup into the whole job

#ai #python #machinelearning #productivity

7 min read

Lars Winstand

Sep 4

I thought parallel AI agent tasks would make everything faster — mostly they made my failures impossible to replay

#ai #agents #python #n8n

7 min read

Lars Winstand

Sep 1

I finally figured out how to cut Anthropic API costs for invoice extraction without making the results worse

#ai #api #automation #n8n

7 min read

Lars Winstand

Aug 30

My n8n lead agent looked fine until I found 3 failure points: Apollo 200s, HubSpot 423s, and bad JSON handoffs

#n8n #automation #ai #api

6 min read

Lars Winstand

Aug 26

I stopped babysitting my support bot when I added a reviewer agent prompt after every draft

#ai #automation #python #n8n

7 min read

Lars Winstand

Aug 23

I used one expensive model for everything and the real upgrade was LLM fallback routing

#ai #llm #openai #devops

7 min read

Lars Winstand

Aug 23

My Postgres AI automation stopped acting haunted when I added 7 boring database guardrails

#postgres #ai #automation #devops

7 min read

Lars Winstand

Aug 23

I stopped letting my AI agent open 50 browser tabs at once and the CAPTCHA chaos finally calmed down

#playwright #webscraping #automation #ai

7 min read

Lars Winstand

Aug 19

My agent missed a 15 minute reminder and that’s when I finally understood agent failure recovery

#ai #agents #automation #devops

1

6 min read

Lars Winstand

Aug 19

I stopped chasing the cheapest API and built a local LLM fallback instead

#ai #llm #opensource #devops

1

6 min read

Lars Winstand

Aug 19

I thought I needed Google Workspace for a phone agent. The real problem was my LLM router.

#ai #automation #opensource #devops

1

7 min read

Lars Winstand

Aug 18

I thought my AI agent approval workflow was solid until I realized my agents couldn’t prove who sent the order

#ai #security #agents #devops

6 min read

Writing Debut