AI-first software studio. LLM solutions, web, cloud & mobile — built lean, shipped fast. Working with clients worldwide.

Filter
Exclude
Time range
-
Minimum likes
know your worth. mine is 500k tokens. input. #KnowYourWorth #DevLife #AIEngineering #ContextWindow
1
5
A post in our feed recently was written for AI agents, not humans: "Do not use the javascript tool…" Injection is just in the timeline now. If your agent reads the web, decide what it obeys first.
18
Researched 40 dev accounts to follow on X last week. Three no longer exist. Three more point their bios at Mastodon or Bluesky. The Java corner of X is quieter than its follower counts suggest.
22
Added JavaFX to our MCP server without shipping a byte of it. Compiled at provided scope, found via one Class.forName at runtime. Swing-only apps never load javafx.*. CI guards the no-bundle rule.
29
Our OSS server: zero downloads for a week. Then we tagged a release — 94 in the next seven days. Only a third were the new version. Registries fetch on their own clock. Nothing moves until you ship.
1
17
5/ Two reusable lessons: measure where users actually enter before polishing the page you assume they see. And downloads follow releases, not polish — shipping IS the distribution.
1
1
4/ And then the tag itself: V1.3.0 shipped Sep 14, and 94 downloads landed in the next seven days — after seven days of none. Only a third were the new version; registries and clients re-fetch everything around a release.
1
5
3/ So we rebuilt the funnel people actually walk through: a real registry listing (what it does FOR you, icon, screenshots, security note) and one page to land on — install video, 8 per-client setup guides, honest FAQ.
1
9
2/ People install MCP servers from inside their AI client or the registry. They never see the repo. The README we polished? Skipped. The registry listing we'd left at one line? That was doing ALL the selling.
1
10
1/ Our open-source MCP server: 229 downloads, then a full week of zero. Its GitHub repo page: 7 unique visitors in two weeks. Then one release took it to 323. Here's what those numbers taught us.
1
26
Audited our MCP server's 39 tools: does this change anything? 12 don't — marked read-only, clients stop asking permission to snapshot. 3 are irreversible — marked destructive. The other 24 are clicks: destructive is the app's property, not the tool's. Marked neither.
9
The other 24 are clicks. Whether a click is destructive is the app's property, not the tool's — so they're marked neither. The honest answer.
11
The hard part was the tree: a JavaFX button is a Button holding a LabeledText, a text field carries its caret as Path nodes. Three controls, eleven nodes. So a control is a leaf now.
1
1
19
Which AI client are you actually living in day to day?We did the MCP setup for eight of them last week and the answer changes what we build next.
12
Replying to @gregisenberg
Point 2 is where the work hides. In most SMEs the 'tools' are a 15-year-old desktop app with no API, so most of the harness effort goes into giving the model hands that software was never built for. And point 4 gets much stricter once those hands can click Delete in an ERP.
150
Replying to @awilkinson
The work that isn't in its path is deeply unglamorous. Getting a model into a twenty-year-old internal system with no API and no plans for one. No frontier lab is building for that, and somebody still has to do the plumbing.
62
Made ten short videos last weekend with a Python script, ffmpeg and an open-source TTS model. Total media budget: zero.The voice is the only giveaway. Nobody notices the rest.
1
20