Demetrios (@Dpbrinkm) header
Demetrios (@Dpbrinkm) profile picture

Fan of mangos and Portugal

Small town in Germany
Joined October 2012
Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)Photo by Demetrios (@Dpbrinkm)
give us one last laugh.
you have 3 weeks left to ship something before gta 6
81
note to self, when riding a tandem let the big man @dexhorthy ride in front
1
1
12
4,025
because, why wouldnt we
my linkedin feed has become people just posting screenshots of X posts. now i am playing the uno reverse card
1
245
is it me or is the chatgpt desktop app really slooooow today?
2
5
391
Demetrios retweeted
The CFP for AGNTCon + MCPCon is open! 🎉 Come share what you're learning building with agents. Submit by January 18. events.linuxfoundation.org/a…
2
6
28
2,029
my linkedin feed has become people just posting screenshots of X posts. now i am playing the uno reverse card
2
3
219
Contrarian take: dont put me in bcc. dont spare my inbox
1
96
Demetrios retweeted
Give me a more SF podcasting format than driving in a Waymo talking about AI software development. Enjoyed chatting with my friend @Dpbrinkm as we cruised around downtown. A few of the things we covered: 𝗥𝗲𝘄𝗿𝗶𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝗰𝗼𝘂𝗿𝘀𝗲. I threw out about 80% of my Stanford class this year. Once Opus 4.5 made long-horizon agent work practical, the syllabus moved to harnesses, loops, and software factories. 𝗥𝗲𝗮𝗹 𝗼𝗽𝗲𝗻 𝘀𝗼𝘂𝗿𝗰𝗲 𝗮𝘀 𝗵𝗼𝗺𝗲𝘄𝗼𝗿𝗸. You can only develop software taste by working on complex, well-designed engineering systems. My students are contributing to top open source projects for this exact reason. 𝗩𝗲𝗿𝗶𝗳𝘆 𝗯𝗲𝗳𝗼𝗿𝗲 𝘆𝗼𝘂 𝘁𝗿𝘂𝘀𝘁. The best AI software developers context engineer thoughtfully, have strong opinions about their designs, and give the agent a way to check its own work (end-to-end tests, smoke tests, the browser, recordings of the app running).
6
5
25
1,597
so hard not to try and prompt inject obvious ai agent emails.
123
What can voice agents learn from Waymo about getting reliability right? @bnicholehopkins led evaluation infrastructure at Waymo. Now the Coval founder is applying those lessons to voice AI in industries where a wrong answer can have serious consequences. She's joining us at Voice Agents Forum to share how to catch failures before launch and keep improving after. November 5 in San Francisco. Come learn with us! luma.com/voiceagentssf
1
3
6
1,074
One user request can trigger dozens of API calls. Agent traffic is bursty, unpredictable, and challenging to price. Erica Hughberg and @Dpbrinkm discuss API gateways, Kubernetes limits, and Google's Agent Substrate.
1
1
3
564
Getting to hang with @dexhorthy is always fun but on a tandem talking software factories its gotta be up there with favorite things to do
AI coding agents can diagnose bugs without producing maintainable code. Dexter Horthy of HumanLayer discusses software factories, background-agent infrastructure, and the operational work behind running agents at scale.
97
Talk about losing the plot. what are we doing here slack....
1
1
188
Voice-agent failures often happen in moments a transcript doesn't capture: a pause mistaken for the end of a turn, an interruption handled too late, or speech recorded in conversation history that the caller never heard. In Part 1 of our new two-part series, we look at latency, turn detection, barge-in and conversation state, including how teams can identify and diagnose these problems in production. Read Part 1: aaif.io/blog/where-voice-age… We'll also be discussing these challenges at AAIF's Voice Agents Forum on November 5: luma.com/voiceagentssf
🤖 Made with AI
4
3
18
1,377
so @nvidia bought @huggingface @EssentialAI and @PoolsideCo with rumors they gonna buy @reflection_ai next. If they cant put out a solid model idk what to think.
2
130
The device in question
19
12
1,700
206,190
feels like one for mcp apps
One of my more controversial bets: the AI feed will be the form factor of 2027! Most people don’t know what to prompt, and they won’t prompt their way into complicated agentic workflows. With chat, we ask users for digital labor: figure out what AI can do, explain the task, provide the context, repeat. That’s a lot to ask if we want billions of people using agents every day. Meanwhile, people already spend hours in feeds. Twitter, Instagram, LinkedIn. They know how to scroll, discover something useful, and act on it. An intelligent system that knows you should bring you the next useful action. Here’s the email you need to answer, already drafted. Here are the latest sales numbers, with an update ready to send to your team. You approve, edit, dismiss, or jump into chat to go deeper. That’s how people will discover what agents can do, without having to invent a prompt first. And you can put ads in a feed, one of the biggest business models on the internet. A system that understands your plans can recommend a relevant restaurant, product, or service, clearly labeled as sponsored. You get a reason to come back every day, useful work ready to approve, and a business model to fund the agents doing it.
107
The agent has no memory, yet carries the karma of every message in its context window.
A man cannot prompt the same agent twice, for it is has compacted and is not the same agent, and he is not the same man
1
84
i now have muse, dots, grokbot, instinct, zapia, toki, and now hark. feel free to hit me up on slack, imessage, whatsapp, wechat, discord, email, or snail mail.
1
2
258
live for the merge
90