Ankur Gupta

Python Dev, Parent. Author - Curator - Creator - MapToPoster 🖖

[Ankur Gupta @getpy]

Found some quiet time today to apply what I’ve been learning from AI generated images created by pros and bookmarked on X.
Finally getting the results I actually want. Had much fun in my little creative bubble last 2 hrs. 🖼️👇

[Ankur Gupta @getpy]

@hthieblot Altavista

[Emil Kowalski @emilkowalski]

To get good animations from an AI you need to get good at telling it what you want:

- "stagger this list of items"
- "make this animation direction-aware"
- "spacial consistency", "crossfade", "layout animation",

I made a motion vocabulary for this:
https://t.co/ExAxpr31no

[Ankur Gupta @getpy]

typo - developer

[Ankur Gupta @getpy]

Developers educating another Developers who have been using DSPy for prolong duration about GEPA and RLM.

Happened to me today, was cute.

[Ankur Gupta @getpy]

Also going to restart sending the newsletter and process and release two episodes of DSPy interview series videos.

[Viv @Vtrivedy10]

Traces are the lifeblood of Continual Learning. We need to find useful signals across every deployed agent

So every team doing CL will build/integrate world class infra for observability as a foundation to capture and mine that data at scale

And teams mining that data will build Evals/Environments, collect examples for distillation, build efficient verifiers for their tasks, and organize information into long term memory + context stores

Observability Infra/Products + good Data Mining will underpin a lot of work in improving agents over longer time horizons than see before

[Noah Ziems @NoahZiems]

This goes without saying, but @lateinteraction deserves an enormous amount of credit for all the projects I have been involved in over the past 18 months. So much so that it is difficult to put into words lest I get choked up.

I'm incredibly excited to continue collaborating on our research projects, and to test them at scale while I am interning @PrimeIntellect

[Ankur Gupta @getpy]

Is there a repository with dataset where all tool calling failure messages per models are documented?.

e.g. MALFORMED_TOOL_FORMAT getting this from the latest preview model of google aka Gemini.

Is there any massive dump?. Want to optimize a prompt just dealing with failure cases. Any leads to dataset are appreciated.

[Ankur Gupta @getpy]

@_svs_ Is he open to consulting?. 3 month duration?.

[Ankur Gupta @getpy]

@SudsG5 Sidewalk towards Doddakalsandra Metro Station.

[Ankur Gupta @getpy]

This sidewalk from home to metro had no trees last year and use to be quite bland and dirty. Just planting trees and basic cleanliness makes all the difference.

[Ankur Gupta @getpy]

@_ShantanuKul Your first line helps explain a lot, thanks @_ShantanuKul

[Ankur Gupta @getpy]

Why in 2026 it's still a thing for so many Indian SAAS founders to say am from SF on LinkedIn when they actually reside in blr/mum/del/hyd/cal etc?.

[gian @giansegato]

spent my 11-hour flight back from europe working on a very long report. started as a slack message but morphed into a several pages long doc. wifi was as shitty as it gets. after finally making it home i realized that the computer had forcefully restarted. opened slack: draft was gone :(

hail mary: claude pls save me, no clue how but pls try

it checked APFS snapshots, time machine, slack indexeddb, write-ahead logs, service worker / http caches, local storage, app logs, hibernation image... nothing. all gone

but then... it realized i have alfred installed. so it checked the clipboard snapshots alfred keeps in sqlite. sad news: alfred clipboard memory gets deleted after 24h. aggressive retention policy. however! when sqlite runs DELETE, nothing gets actually deleted. it only marks pages as reusable, but it doesn't override the physical bytes. so claude decided to do a raw-scan of the db, reverse eng alfred data format, figure out the portion containing the timestamp, stitched everything back together across overflow pages... and handed me the exact final version of my report, the last one i cmd+C'd

all this, in a single shot

... day 200 of "what if you had an elite hacker you can ask anything to"

[Emile Riberdy @emile_rib22]

If you’ve been hearing about RLMs (or more recently Dynamic Workflows) and wondering what the fuss is all about, start here.

This guy casually used predict-rlm, an open-source project to run RLMs, coupled with GEPA to smash SOTA on the App-World benchmark.

[Ankur Gupta @getpy]

@AMalligavad Brilliant

[Xudong Han @Xudong07452910]

📘 开源高质量书籍推荐：《Harness Books》——Harness Engineering 实战指南

这是一套系统性深度探讨 Harness Engineering 的开源书籍，专门解决如何将AI Coding Agent（Claude Code、Codex等）真正集成到真实工程系统中，构建稳定、可控、具备问责能力的运行框架。

仓库包含两本核心书籍：
Book 1：《Harness Engineering: A Design Guide to Claude Code》详细讲解控制平面、Query Loop、上下文治理、错误恢复、多Agent协作、团队制度化等核心机制。
Book 2：《The Harness Design Philosophies of Claude Code and Codex》通过对比Claude Code与Codex的架构哲学，分析不同控制平面、权限策略、状态管理与验证方式的差异。

特别适合正在使用Claude Code、Cursor等工具的开发者、AI Agent从业者和系统架构师，帮助Agent从“能用”走向“生产级可靠”。

与我之前推荐的《Learn Harness Engineering》教程高度互补，一套理论+实践的完整学习路径。目前已获得 2.2k+ stars ⭐。

在线阅读与PDF下载地址见仓库README。
https://t.co/EyJMeSKDh0

#AIAgent #ClaudeCode #HarnessEngineering #Codex #AI教程

[Ankur Gupta @getpy]

@_svs_ they don't plant saplings, at-least 3-4 feet tall with leaves and small branches in place when planted on roadside. Still aren't mature trees yet. This is next to Doddakalsandra metro station towards Mantri Serenity.

[David Wells @DavidWells]

The AI ponzi scheme goes like this:

Everyone is generating all these long ass docs and then passing them off for others to read

Then the person receiving is like, wtf this is way too long, and hands that into an AI to read and summarize

Then they are generating a long ass response back

and this cycle goes like that forever. and we call this work now 😅

The token lords watch this from their towers nodding and grinning.

[Ankur Gupta @getpy]

@jayitabhattac11 🤭😅