←
AI.Engineer3,857 posts
AI.Engineer avatar

AI.Engineer

@aiDotEngineer

We run the highest signal technical conferences in the industry and publish it all for free online for over 2 million AI Engineers a month

◷ Joined March 2021▧ 980 Photos and videos
dom hofmann avatar
⟳ AI.Engineer retweeted
dom hofmann@dhof·Aug 5

an assistant for most people probably starts as a feelings manager more than a task manager stress comes from both factors and mindset

VerySmallWoods avatar
⟳ AI.Engineer retweeted
VerySmallWoods@verysmallwoods·13h

看了 @badlogicgames Mario Zechner 的几场访谈,Pi SDK 和 Pi Durable 的定位和适用场景清晰了许多。 简单总结一下: - Pi SDK:把 coding agent 的会话、工具和扩展嵌入自己的应用。 - Pi Durable:为长期运行的 agent 提供任务检查点、崩溃恢复、调度和共享状态。 日常的 agentic app 开发中通常选其一。 推荐按这个顺序看: Building pi in a World of Slop:从 05:24 看,理解小核心、扩展和工作流控制权。 youtube.com/watch?v=RjfbvDXp… Building Pi,Pragmatic Engineer:从 37:35 看,理解模型接口、agent loop 和 coding agent 的关系。 youtube.com/watch?v=n5f51gtu… Understanding Pi Durable:从 09:41 看,Mario 直接解释为什么已有 SDK,还要另做 Durable。 youtube.com/watch?v=7mBQknCn… Pi Durable 维也纳演讲:从 03:27 看,深入任务状态机、恢复和工具重放的边界。 youtube.com/watch?v=EreOxs9V…

Hux avatar
⟳ AI.Engineer retweeted
Hux@huxlab·Nov 17

“所有代码都是垃圾(slop),除了我的。” 在 AI Engineer Code Summit 上,SpaceX / xAI 的 GrokBot 工程师 Lauren Tan 抛出这句暴论后,直接甩出一组硬核数据: 上个月她一个人往生产环境合了 3,000 个 PR(全团队 1.1 万个),回滚率只有 0.5% 左右。 大多数人还在死磕 Prompt,她却靠一套反直觉的工程化打法,让 Agent 在她睡觉时自己写代码、自己验、自己合: 1. 代码库本身才是最大的输入 Agent极其擅长“照猫画虎”。与其在文档里苦口婆心写规矩,不如直接把容易出错的路径物理封死:比如产品代码里禁用 useEffect、跨进程 import 直接报错、注释几乎全禁。 2. 每次 Agent 犯错,改成 Lint 而不是改 AGENTS.md 自然语言规则(prose)是最脆弱的。每次纠正完 Agent,她都会把教训固化成确定性的静态检查或架构约束,让同类错误在提交前就绝无可能发生。 3. 用运行时验证让 Agent “自证清白” 写完不算完,强制 Agent 自己跑测试、抓截图、录视频,甚至上形式化验证,拿实打实的运行结果证明功能没问题。 4. 搭建自下而上的“信任金字塔”代码库约束 > 静态分析 > 运行时验证 > 人眼 Review。 把确定性交给工具,把人的精力留到最后一道关口。 很多团队花几个月试错才摸到的门道,她用 17 分钟就讲透了。 真正拉开人机协作差距的,从来不是模型有多聪明,而是你有没有把环境改造成“最短路径就是正确路径”: 把约束写进代码里,而不是写在文档里。

17:18
AI.Engineer avatar
AI.Engineer@aiDotEngineer·13h

Code getting cheaper makes judgment more valuable. What deserves to exist is still an engineering decision. Good takeaway from @RamiSyriani.

Wrapped up @aiDotEngineer code summit yesterday! Spent great time meeting passionate, amazing people and hearing from speakers on building software factories. A couple of takeaways: stay ambitious, and just because code is cheap now and you can build it doesn’t mean you have to.

Post attachmentPost attachmentPost attachment
Roshan Chandna avatar
⟳ AI.Engineer retweeted
Roshan Chandna@roshanchandna·21h

At @GoogleDeepMind, @melvinjohnsonp says they switched from grading coding agents on unit tests to rubrics that also score style and maintainability, like a strict code reviewer

0:24

An agent can pass the tests and still miss the point. @melvinjohnsonp of @GoogleDeepMind on grading agents like a strict reviewer: correctness, style, maintainability, and penalties for gaming the system. From AI Engineer Code Summit.

1:06
Allie Howe avatar
⟳ AI.Engineer retweeted
Allie Howe@vtahowe·19h

It was great talking to @TejasKumar_ at @aiDotEngineer CODE about his workshop and Daiku Daiku is a personal agent harness and he told me it filled out Shark Tank application on his behalf and got him on the show 🤯 Go checkout daiku.ai and this new pod

Roll Your Own Agent Harness: Building Daiku and Personal Agents with Tejas Kumar "Let me send a prompt here, 'Delete everything.' That's what I did on stage. And it actually just did it." @TejasKumar_, AI Engineer and author of Fluent React, talked to us at @aiDotEngineer CODE in SF about his AIE CODE workshop, building agent harnesses from scratch, and why "can the agent call this tool" and "should it call this tool right now" are two different questions. We get into: > Why banks, insurers, and hospitals are building their own harnesses instead of using off-the-shelf coding agents > The guardrails that came after rm -rf: allow lists, ask-first tools, and Face ID before sensitive actions > Getting the right human in the loop, not the whole team > Context caching, the piece most self-built harnesses are missing > Daiku, a personal agent harness that isn't locked to one model vendor TIMESTAMPS (00:40) The workshop: from an empty while loop to a coding agent harness (01:45) Rolling your own harness for air-gapped enterprise infrastructure (02:40) Why regulated industries can't just use off-the-shelf harnesses (03:40) Can you call the tool vs should you call it right now (04:10) "Delete everything" on stage, and the guardrails that came next (04:50) The right human in the loop (05:50) Context caching and why self-built harnesses overpay for tokens (07:20) The Resy incident: an agent that tried too hard (07:55) AAuth and budgets for downstream resources (09:50) Keeping credentials out of the agent and the deployment (11:15) The power plug you can pull on a rogue agent (12:05) Daiku: one computer every one of your agents can drive (13:05) The agent that signed Tejas up for Shark Tank (15:45) Switching models mid task when you run out of usage

17:35
AI.Engineer avatar
AI.Engineer@aiDotEngineer·16h

Agents don't search like people. @jpinner of @ExaAILabs compares short human searches on @Google with agent queries full of dates, versions, locations and constraints. When a query asks for companies outside the US, returning a US company isn't close enough.

1:12
AI.Engineer avatar
AI.Engineer@aiDotEngineer·22h

Prefer reading to watching? 50 talks from AI Engineer Code Summit, now in writing: illustrated essays, full transcripts and links to watch. Pick a topic. Find the idea you want to try. ai.engineer/code/2026#talks

AI.Engineer avatar
AI.Engineer@aiDotEngineer·Nov 17

"If you want good taste, you have to eat." @jxnlco of @OpenAI on studying good work, trying other people's tools, and learning to explain what makes something good. From AI Engineer Code Summit.

1:35
AI.Engineer avatar
AI.Engineer@aiDotEngineer·Nov 16

Thousands of tool definitions in the context window? @mitsuhiko of @earendilworks makes the case for giving agents JavaScript to run, so they can compose tool calls into a workflow. From AI Engineer Code Summit.

1:59
Matt Pocock avatar
⟳ AI.Engineer retweeted
Matt Pocock@mattpocockuk·Nov 16

.@poteto's talk at AIE Code is excellent. If I had to pick one still from that felt genuinely original, it would be this one: "nuance tends to lead to slop" To increase quality, design API's and automated checks that remove the nuance from agents' decisions. - Ban useEffect - Ban 99% of explanatory comments (they just lead to workarounds) - Create API's that hide complexity from agents Full talk below.

Post attachment