📢 Die Ära der 𝐈𝐧𝐝𝐢𝐯𝐢𝐝𝐮𝐚𝐥𝐬𝐨𝐟𝐭𝐰𝐚𝐫𝐞 beginnt am Küchentisch

Tobias Boehm hat vor vier Tagen eine klasse 𝐋𝐞𝐫𝐧𝐚𝐩𝐩 für seine Tochter veröffentlicht. Bunt, ansprechend gestylt, direkt zugeschnitten für die Zielgruppe: eine Person. Für ein einzelnes Ziel: Sicherstellung guter Noten. (Was ist eigentlich rausgekommen?)

Der Post sprach mich direkt an, da ich in der Vergangenheit ebenfalls versuchte, abseits bestehender Lernsoftware etwas für die Kinder zu erstellen: manchmal mit mehr Publikumserfolg, meist mit weniger. 𝐆𝐚𝐦𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 ist definitiv der richtige Ansatz, um die jüngere Zielgruppe zu interessieren. Bei mir schwang auch der Aspekt mit: Zeig den Kindern einmal direkt, was mit etwas Einsatz, Kreativität und Programmierkenntnissen möglich ist. Vielleicht zieht es sie ebenfalls in die richtige Richtung. Vielleicht ist man auch manchmal überinvestiert in den Erfolg der Kinder – das mag sein. Aber warum auch nicht?

Aber zurück zum Thema: Ich finde das Projekt richtig gut und es zeigt, dass man mit agentischer Unterstützung und dem passenden Software-Know-how individuell zugeschnittene Software-Produkte erstellen kann. Hochqualitativ (siehe Tests und Pipelines) und in kurzer Zeit. Dass es als 𝐎𝐩𝐞𝐧 𝐒𝐨𝐮𝐫𝐜𝐞 mit passender Lizenz veröffentlicht wurde, erlaubte mir ebenfalls, einen winzigen Teil beizutragen – direkt nach Review durch Agenten gemergt. Super.

🦄 𝐌𝐚𝐭𝐡𝐞𝐥𝐞𝐫𝐧𝐞𝐧 𝐝𝐚𝐫𝐟 𝐒𝐩𝐚ß 𝐦𝐚𝐜𝐡𝐞𝐧! 🎉

Link zum Ursprung: https://lnkd.in/p/eRMQqGp9

𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐈 𝐈𝐬 𝐤𝐢𝐥𝐥𝐢𝐧𝐠 𝐭𝐡𝐞 𝐭𝐢𝐜𝐤𝐞𝐭

There is a big question: could it be that agentic AI will be the death of collaboration software, or dashboards like #Jira and #GitLab, or even #DevOps, in the sense that nobody needs a ticketing system anymore?

Even when you work professionally with AI, you can move so fast that you can get things analyzed, checked, and also weighed. In terms of: what is the best option? Then you do the implementation, do the vetting of the implementation, and merge it.

So you move so fast that the bureaucratic overhead of maintaining a ticket system, maintaining ticket states, and also creating official merge/pull requests becomes quite drastic.

And let’s put it like this: when you use frontier models, I haven’t seen much of a quality difference. If you have proper instructions and tooling-guardrails, you can also get these things done entirely locally with AI. So there is no need to first send something to some kind of remote repository, then create a merge request or pull request there, do the review by a second person (four eyes ..), have back-and-forth discussions, run the CI, and then merge it.

I mean, even regarding that human review thing, I’m not sure whether this still fits. The review by a second critical agent can be done locally. Running a pre-push CI should be done locally anyway, because then you also save the time required to execute things remotely and seeing it fail there. Of course, remote CI adds an additional safety layer because things are checked independently and 𝐝𝐨𝐜𝐮𝐦𝐞𝐧𝐭𝐞𝐝.

But I have to say, on private projects, I stopped doing tickets.

Everything can be handled within 𝐌𝐚𝐫𝐤𝐝𝐨𝐰𝐧 𝐟𝐢𝐥𝐞𝐬 𝐢𝐧𝐬𝐢𝐝𝐞 𝐭𝐡𝐞 𝐫𝐞𝐩𝐨𝐬𝐢𝐭𝐨𝐫𝐲, because then you have one place where everything is 𝐜𝐞𝐧𝐭𝐫𝐚𝐥𝐢𝐳𝐞𝐝.

Also I have to say, with agentic AI it comes down to one thing in terms of implementation: is this really something you want to 𝐜𝐫𝐞𝐚𝐭𝐞 𝐧𝐨𝐰, or do you want to do it in the 𝐟𝐮𝐭𝐮𝐫𝐞? “Future” means: when there is an opportunity and there is time. No more: is this now important or that? Eversthing important now, the rest later. Capacity scheduling solved.

The thing is: what is your view on that topic?

That’s my point of view. Picture totally unrelated and from a lunch-break wall. Enjoy autumn 🍂 🎃

𝐭𝐥;𝐝𝐫

idea → ticket → refinement → assignment → implementation → PR → review → revision → CI → merge
can increasingly become:
intent → agentic work session → verified commit

Chart of the day: 𝐦𝐨𝐝𝐞𝐥 𝐮𝐬𝐚𝐠𝐞 𝐢𝐧 𝐒𝐞𝐩𝐭𝐞𝐦𝐛𝐞𝐫

This shows a breakdown of the currently ending month of September (in 2026). Which models were used, with vendor and #tokens burned, and an estimate of what this usage would have cost via the API (most of it was covered by subscriptions). As you can see, it is a mix of the top dogs of the Western hemisphere as well as self-hosted #Qwen-based #Ollama runs. Since for self-hosting I could only make up numbers for the dollar value, I kept it with no mapping at all, just show the token usage as number.

The screenshot shows a small part of tokenUsage2 – which is a TUI tool I developed some weeks ago to break down token usage with ad hoc charts. Updated every few milliseconds. Not like the HTML generator I showed some weeks ago. 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬 are fun – but you also need the 𝐜𝐨𝐦𝐩𝐮𝐭𝐞 to get them done. And some 𝐨𝐫𝐜𝐡𝐞𝐬𝐭𝐫𝐚𝐭𝐢𝐨𝐧 skill.

Find it on GitHub. I’ll spare you the URL here, since I know that almost nobody clicks the outbound URLs 😉

𝐖𝐡𝐚𝐭 𝐚𝐫𝐞 𝐲𝐨𝐮𝐫 𝐧𝐮𝐦𝐛𝐞𝐫𝐬? 𝐂𝐚𝐫𝐞 𝐭𝐨 𝐬𝐡𝐚𝐫𝐞?

#anthropic #openai #codex #claudecode #agenticengineering

#

Last week I saw here on LI a post by a father where his 9-year-old son created his own Minecraft clone with agentic AI. Even for the love of the gods, I can’t remember who it was. (If someone knows, ping me.)

While reading, I thought: wonder how much the father helped. And Minecraft is so popular. Even with tons of mods, this is a well-documented topic.

Yesterday, between breakfast and getting ready for an excursion, I asked my oldest child if she wants to create her own game and has ideas. Yes… she said… I wasn’t even ready with preparing a repository when she already went on and on and on… and I thought, omg, this will not work out. What have I stirred up? 🫣

One hour of chores later, we returned. And it was playable. A friendly, colorful cat adventure. With reward mechanisms. Great :3

My only input was my “dark factory prompt” and that it shall use Babylon.js (good experiences with that for 3D graphics in the browser). The rest was done by a slightly overexcited product manager and AI ™

Outcomes:
* I overestimate my necessary input for agentic AI
* I can’t publish a fully working and polished game, because it is hers
* only our inspiration is the limit

𝐂𝐥𝐚𝐮𝐝𝐞 /𝐢𝐧𝐢𝐭 – 𝐮𝐬𝐞𝐥𝐞𝐬𝐬!?

Last week I attended an 𝐢𝐥𝐥𝐮𝐦𝐢𝐧𝐚𝐭𝐢𝐧𝐠 𝐰𝐨𝐫𝐤𝐬𝐡𝐨𝐩 by Zartis, which brought knowledge about Claude Code and, on the other hand, the chance for networking. My table neighbours and I helped each other with some challenges, and then the question came up: does the /init command actually help? The common understanding, because it is advertised like this: yes. If you have a legacy project and want to give the agent a head start. 𝐁𝐮𝐭 𝐡𝐚𝐬 𝐚𝐧𝐲𝐨𝐧𝐞 𝐞𝐯𝐞𝐫 𝐦𝐞𝐚𝐬𝐮𝐫𝐞𝐝?

I did some checks on available research on the fly. But the outcome was ambiguous.

Therefore, here is now the summary of 𝐦𝐲 𝐦𝐞𝐭𝐚-𝐬𝐭𝐮𝐝𝐲: as a 5-page carousel. I’ll add a 𝐭𝐡𝐫𝐞𝐞𝐩𝐚𝐠𝐞r 𝐰𝐡𝐢𝐭𝐞𝐩𝐚𝐩𝐞𝐫 𝐚𝐬 𝐭𝐡𝐞 𝐟𝐢𝐫𝐬𝐭 𝐜𝐨𝐦𝐦𝐞𝐧𝐭. AI-co-authored, of course, but I spent several sessions checking different angles and the six publicly findable studies.

𝐎𝐮𝐭𝐜𝐨𝐦𝐞𝐬:
* no reliable gain!
* it costs more: 20% per task in the ETH Zurich study
* random rules score as well as expert ones: so you don’t need to give someone their comment <– check: commit to get their master claude.md 😉
* rules get followed and named tools are used more often
* writing down the agent’s mistakes worked: especially for smaller (35B) models – self-hosters, listen up
* agents.md/claude.md files only grow, almost never shrink!

And I find this underlines my view. 𝐓𝐡𝐞 𝐟𝐢𝐞𝐥𝐝 𝐦𝐨𝐯𝐞𝐬 𝐬𝐨 𝐟𝐚𝐬𝐭 𝐭𝐡𝐚𝐭 𝐚 𝐥𝐨𝐭 𝐨𝐟 𝐬𝐨𝐥𝐝 𝐭𝐫𝐮𝐭𝐡𝐬 𝐚𝐫𝐞 𝐛𝐚𝐬𝐞𝐝 𝐨𝐧 𝐬𝐮𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞 𝐞𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞𝐬.

I personally almost never used the /init command ..

carousel: LI

whitepaper: paper

𝐀𝐠𝐞𝐧𝐭𝐖𝐡𝐢𝐥𝐞𝐓𝐫𝐮𝐞 has got the 𝐰𝐡𝐢𝐩 now!

A decade ago whip apps for your favorite smartphone were the hit. Using the acceleration sensor, they played a whipping sound when you made the movement while holding your phone. A team lead at that time jokingly always used it (nobody was harmed …).

Poetic afflatus: why not add this to Agent While True? It is a machine talking to other machines (means: agentic harnesses in some terminals).
And here it is: with a simple press of 𝐖 you’ll unleash hell.
A selection of twenty comments in the style of 𝐓𝐡𝐞 𝐎𝐟𝐟𝐢𝐜𝐞/𝐒𝐭𝐫𝐨𝐦𝐛𝐞𝐫𝐠 are sent via #DBus to the open sessions of #ClaudeCode or #CodexCLI that have work left. And you see a nice cracking animation (#ASCII of course).

𝐃𝐢𝐬𝐜𝐥𝐚𝐢𝐦𝐞𝐫: just because I have added this fun-feature to the TUI does not mean I endorse questionable motivational methods or intimidation. A workplace, like every professional and human interaction, should be based on respect!

To address the HR-fellows: how do you like that new feature? 😉