Notes from the work

One email when a tutorial or migration write-up goes live. Nothing else, and an unsubscribe link in every email.

LeanZero

Two people in Romania doing Atlassian migrations, Forge apps and practical AI work for teams that would rather talk to the person doing the job. Most of what we learn ends up on this site.

Services

  • Atlassian migrations
  • Atlassian FastShift
  • Atlassian maintenance
  • Forge app development
  • AI development consultation

Topics

  • Jira
  • Jira Service Management
  • Confluence
  • Bitbucket
  • Atlassian Forge
  • Cloud Migration
  • Local AI
  • AI Coding
  • Certifications
  • Atlassian Team
  • All topics

Company

  • Blog
  • Tutorials
  • Contact

Community

  • Join Discord
  • Support this site

© 2026 LeanZero. All rights reserved.

Privacy policyTerms of service24/7 SupportTrust centerSecurityDPALegal notice
  1. Home
  2. Blog
Knowledge Base

Atlassian insights & tutorials

Real Atlassian solutions to real problems — no fluff, no SEO spam.

Read the latestFollow the tutorials
JiraConfiguration, workflows and fixes from real projectsLocal AIModels and agents running on your hardwareAI codingCoding tools and workflows tested in practice

Articles 15

All topicsAI CodingAtlassian ForgeAtlassian TeamBitbucketCertificationsCloud MigrationCogniRunnerConfluenceCost SlasherJiraJira Service ManagementLeanZero ManagementLocal AISentinel Vault
I taught a 27B model to write Forge apps. As far as I can find, nobody had done that before
ArticleLocal AIAtlassian Forge

I taught a 27B model to write Forge apps. As far as I can find, nobody had done that before

Qwen3.8-27B knows almost nothing about Atlassian Forge: ask it for an app and it invents a manifest format and imports packages that do not exist. Over four days on one Mac Studio I trained it until 14 of 25 manifests passed Atlassian's own validator and 12 of 25 apps compiled, from zero. The weights are public. Here is what it is, what it is not, and why I think it is a first.

Mihai PerdumSep 10, 202611 min read

More articles

The best 4-bit format in MLX is the one its own converter sets up to lose
ArticleAI Coding

The best 4-bit format in MLX is the one its own converter sets up to lose

NVFP4 has a second scaling level. mlx-lm never sets it, so NVFP4 converts to the worst 4-bit result on the table — worse than a format using fewer bits. Set it and NVFP4 wins. I nearly published this as a defect in NVIDIA's spec; it was my own unit error.

Mihai PerdumSep 1, 20269 min read
What 145 MCP Tools Cost Before the Model Reads Your Question
ArticleAI CodingLocal AI

What 145 MCP Tools Cost Before the Model Reads Your Question

MCP went stateless on 2026-07-28. I probed ten real servers: none implement it. Then I measured the thing that actually costs you — 145 tool definitions, 40,784 tokens, and the 30% that is one server repeating itself.

Mihai PerdumAug 17, 202614 min read
96 GB is 77.76 GiB: the real memory ceiling on an M3 Ultra
ArticleAI CodingLocal AI

96 GB is 77.76 GiB: the real memory ceiling on an M3 Ultra

Metal will not give you the RAM on the box, and the number it does give is not the 75% everyone repeats. I measured the three ceilings on a 96 GB Mac Studio, then measured what modern hybrid-attention models actually spend against them — including a Gemma 4 cache that quietly holds three times its own sliding window.

Mihai PerdumAug 13, 202613 min read
Ling 3.0 Flash won't load on a Mac Studio. Qwen3-Coder-Next does, at 73 tok/s.
ArticleAI CodingLocal AI

Ling 3.0 Flash won't load on a Mac Studio. Qwen3-Coder-Next does, at 73 tok/s.

Ant Group's 124B/5.1B-active hybrid-linear MoE hit Hugging Face on 2 August. The memory arithmetic fits a 96 GB Mac with room to spare, and mlx-lm still refuses it. I counted exactly which tensors block it — 385 of 62,237 — then measured what that sparsity actually buys on the models that do run.

Mihai PerdumAug 5, 202618 min read
MXFP8 vs Q8: 10x the weight error, 1% the perplexity
ArticleAI CodingLocal AI

MXFP8 vs Q8: 10x the weight error, 1% the perplexity

I quantised real Qwen3-Coder weights both ways on an M3 Ultra. MXFP8 reconstructs them about 10x worse than 8-bit affine at identical size — and then costs only 1% perplexity end to end. Both numbers are true, and the gap between them is the interesting part.

Mihai PerdumAug 1, 202614 min read
Previous
Page2of 2
Next

Found my advice helpful?

If I helped you solve something for free, coffee money is always appreciated (but never expected).

Buy Me a Coffee

Join the community

Join the community to ask questions, share your work, and connect with other professionals.

Join Discord

Recent writings

View all
Ricursive Ripple-65B Preview benchmark: 0.028 on Gauntlet, 0.0814 on Forge, plus LFM3-1.7B
ArticleLocal AI

Ricursive Ripple-65B Preview benchmark: 0.028 on Gauntlet, 0.0814 on Forge, plus LFM3-1.7B

Mihai Perdum16 min read
Nanbeige4.3-33B benchmark attempt on a Mac Studio, plus Particle-1 vs DeepSeek V4.1 Flash
ArticleLocal AI

Nanbeige4.3-33B benchmark attempt on a Mac Studio, plus Particle-1 vs DeepSeek V4.1 Flash

Mihai Perdum24 min read
Qwen3.8-Flash-Next on a Mac Studio vs OpenRouter: what it took to reach 0.6932
ArticleLocal AI

Qwen3.8-Flash-Next on a Mac Studio vs OpenRouter: what it took to reach 0.6932

Mihai Perdum20 min read
I ran MiMo V2.6 Flash local on two Mac Studios over Thunderbolt 5 RDMA
ArticleLocal AI

I ran MiMo V2.6 Flash local on two Mac Studios over Thunderbolt 5 RDMA

Mihai Perdum15 min read
Muse Spark Open benchmark: where the open-weight models we ran this month broke as coding agents
ArticleAI Coding

Muse Spark Open benchmark: where the open-weight models we ran this month broke as coding agents

Mihai Perdum20 min read