Daniel Tks

🇪🇺 🇩🇪 | building pareto level agent harnesses for unstructured data processing | cto cambrion

[Daniel Tks @dtks__]

@nikesharora eu people are not stupid. they see that ai is getting good but many are missing a clear on ramp. smth that demos clear value wo the sales bs. if you have proof that stuff is actually working they are willing to compromise on sovereignty

[Daniel Tks @dtks__]

@eurofounder Your vacation is my opportunity

[Daniel Tks @dtks__]

@NielsRogge @Shopify absolutely agree here! prompt opt is going to be part of many enterprise products (inkl. ours!). we see teams at huge orgs starting to self-host. but it's early and the roi is not yet fully visible. interestingly self-hosting is going to be huge in public services...

[Daniel Tks @dtks__]

@tech_optimist I wonder why they get so much hate. Maybe it’s just my bubble…

[Daniel Tks @dtks__]

@itsolelehmann I feel like this is slightly out of touch with medical reality. During the pandemic I had a discussion with a good friend (who is an excelent doctor) that full body mri would solve many problems. She was like nah it wont help because medical diagnosis is much more complex...

[Daniel Tks @dtks__]

@andrewmccalip Postgres all the way

[Pavel Afanasiev @pavelreplies]

I admire Arthur and I wish @MistralAI well, but this communication is incredibly misplayed:

1) the "it's actually le gros chaton" was far more impactful and culturally relevant than any corporate messaging

2) the "we somehow got put in the spotlight" shows that you are absolutely not in charge of the cultural narrative. And the last 5 years show that narrative matters more than anything (more than profitability, more than customers, more than financials) - check SpaceX and OpenAI for notes.

3) long paragraphs about Enterprise and FDEs reaffirm what everyone already thinks (if they think of Mistral at all): You are rebuilding Accenture, not Anthropic. That can be interesting, perhaps even profitable, but it's deeply uninspiring (check Accenture personnel churn rates).

4) because of the above, the conclusion about sovereignty (which is both deeply needed, but also the only actual bullet in the proverbial gun) lands flat.

We (France, Europe, humanity) need you. And we need you to be better, to be more visible, and to be more in control of the narrative. Please. ❤️

[Daniel Tks @dtks__]

@catalinmpit 👋🇩🇪

[Daniel Tks @dtks__]

once you understand how people build agents you realize that this is one of biggest opportunities in ai

[ben hylak @benhylak]

in our blog, we explain an interesting design decision we made: exposing our agent as an MCP tool.

turns out, it has some huge advantages vs. just exposing individual tools

(link below)

[Daniel Tks @dtks__]

@reach_vb @LexnLin Didn’t know you were based in Munich too

[Daniel Tks @dtks__]

i can finally put stuff into production

[Polymarket @Polymarket]

NEW: It’s been revealed that OpenAI’s Codex system prompt includes explicit instructions to "never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other creatures.”

[Daniel Tks @dtks__]

@kentcdodds It's easier to call it "connector" instead of what it is

[Daniel Tks @dtks__]

@signulll heavily agree here!

[Daniel Tks @dtks__]

@climatejulien Is this a direct consequence of polar vortex instability?

[Daniel Tks @dtks__]

@dosco Very glad to have bet the farm on that

[Daniel Tks @dtks__]

@helloiamleonie home built stack with @DSPyOSS and @DBOS_Inc

[Daniel Tks @dtks__]

@benedictk__ Any suggestions in Munich?

[Daniel Tks @dtks__]

@dbreunig It felt like this for earlier versions too

[Daniel Tks @dtks__]

@natolambert Kind of bearish on domain specific fine tuning, but otherwise spot on

[Daniel Tks @dtks__]

@larsencc for people building in the real world every single point is the baseline