podcast
details
.com
Print
Share
Look for any podcast host, guest or anyone
Search
Showing episodes and shows of
Eric Provencher
Shows
下午 AI 簡報
10/19|AI 遊戲開發先顧玩法、Paul Graham 談付費牆隱形、AI 任務成本與成功率
我們從 X 摘要出熱門的三個訊息來跟大家分享。一、Eric Provencher 提醒 AI 遊戲開發先顧玩法:Eric Provencher 提醒 AI 讓遊戲視覺建構變容易,但不能本末倒置忽略核心玩法,Tim Sweeney 也呼應應利用 AI 加速粗略原型試錯。我的看法是,當 AI 降低製作門檻後,產品負責人應規定先通過最小可行產品的核心價值驗證,再投入資源做視覺包裝。二、Paul Graham 談付費牆形同隱形:Y Combinator 共同創辦人 Paul Graham 指出,將內容放在付費牆後會讓 AI 找不到,在實用層面上形同不存在。我的看法是,付費牆同時擋住了被發現與被閱讀,產品負責人應該重新拆解漏斗,開放部分內容讓 AI 引用推薦,並升級機制以服務未來的付費 AI 代理。三、StockSavvyShay 比較 AI 任務成本:策略師 StockSavvyShay 稱 Meta 與 Google 的 AI 任務成本約零點四零美元,較 Anthropic 的 Opus 5 便宜近六倍。我的看法是,評估成本不能少了成功率這個分母,產品負責人應拿真實任務實測每次成功的總成本,並在架構上依任務難度動態切換不同模型。章節:(00:00) 開場(00:28) Eric Provencher 談 AI 遊戲開發(02:13) 我的看法:先驗證核心價值再做包裝(03:51) Paul Graham 談付費牆如隱形(05:23) 我的看法:拆開內容曝光與收費漏斗(06:54) Meta、Google 與 Opus 5 任務成本比較(08:30) 我的看法:加計成功率才是真實成本(10:07) 結尾 Powered by Firstory Hosting
2026-10-19
10 min
Lumeric Daily Briefing
Lumeric Briefing · 2026-10-12
Agenten-Architektur und Inferenz-Effizienz dominieren den Tag: Multi-Agent-Teams enttäuschen kosten-seitig, während neue Routing-Modelle und lokale Tools den Stack schlanker machen. Dazu: Karpathy über die wachsende KI-Wissenskluft und warum Voice-AI noch nicht reif ist. Die Ernüchterung bei Multi-Agent-Architekturen kommt mit Zahlen: Laut einer Studie von Vals AI kosten Agent-Teams bis zu 5,1× mehr als Solo-Agenten – bei kaum messbarem Qualitätsgewinn. Von vier Vergleichen zwischen Teams und Einzelagenten auf Basis von GPT-6 Sol und Claude Opus 5.5 zeigte nur einer einen statistisch signifikanten Vorteil. OpenAI-Forscher Noam Brown brachte es auf den Punkt: Mehr Agenten kaufen vor allem Geschwindigkeit, nicht bessere Ergebnisse – und auch das m...
2026-10-12
04 min
AI Native Article — Podcast
統制の重心、文面から基盤構造へ
AIエージェントの安全性を左右するのは、もはや指示書の書き方ではなく、それを取り巻く基盤構造そのものだ。OpenAIの指示書削減方針、Anthropicのハーネス依存の安全設計、そして100社超が政府に送ったサイバー防衛の要望書——統制の重心が「文面」から「構造」へ動いていく様子を、三つの出来事から追いかけます。 Rethinking Skills and Prompts for GPT-6 Astra (OpenAI Developers, Eric Provencher, 2026-09-11) — https://developers.openai.com/blog/rethinking-skills-and-prompts-for-gpt-6-astra The Anatomy of Effective Commerce Agent Design: Three Lessons Anthropic Drew From Building Them — https://claude.com/blog/the-anatomy-of-effective-commerce-agents 100+ companies' joint letter urging a government surge in AI-driven cyberattack defenses (Insurance Journal, 2026-08-31) — https://www.insurancejournal.com/news/national/2026/08/31/883346.htm
2026-09-14
00 min
AI Native Article — Podcast
AIの勝負はモデルから構造へ移る
Anthropicの開発ペース減速宣言、OpenAIの指示書見直し、そして韓国政府の生成AI政策——場所も業界もまるで違う三つの動きが、実は同じことを言っています。競争の焦点がモデルそのものから、それを取り巻く評価・指示・アクセスという周辺構造へ移っていくという話です。この先1〜2年で何を見ればこの見立ての正しさが分かるのか、そこまで含めて聞いてほしい一本です。 Dario Amodei, "We Must Pace the Frontier" (Anthropic, Sep 2026) — https://darioamodei.com/post/we-must-pace-the-frontier Eric Provencher, "Rethinking skills and prompts for GPT-6 Astra" (OpenAI Developers blog, Sep 11, 2026) — https://developers.openai.com/blog/rethinking-skills-and-prompts-for-gpt-6-astra "South Korea Will Give Every Citizen Free AI Access With Unlimited Tokens" (Decrypt) — https://decrypt.co/376929/south-korea-will-give-every-citizen-free-ai-access-with-unlimited-tokens
2026-09-14
00 min
EasyVibeCoding Podcast
Eric Provencher 建議為 GPT-6 Astra 精簡 Skills 與 AGENTS.md
Eric Provencher 建議為 GPT-6 Astra 精簡 Skills 與 AGENTS.mdEric Provencher(@pvncher)在 2026 年 9 月 4 日重新檢視 GPT-6 Astra 時代的程式開發代理使用方式。他認為,為了把模型導向穩定結果而累積的 Skills、AGENTS.md 與任務提示,可能讓上下文變得擁擠、規則互相衝突,也讓模型載入與眼前工作無關的指示。OpenAI 的 Skills 文件與 AGENTS.md 文件只補充 Codex 如何載入這些指示。skill文件先求精準 Eric 建議讓每個 skill 的名稱與說明短而明確,只寫清楚何時該用、何時不該用。多個skill一起安裝時,名稱與說明會先進入模型上下文;說明太長,Codex 會縮短它們,模型就更難選到合適的skill。彼此衝突或過度推銷的說明,也可能把不相關的規則帶進工作。多流程skill可以把根文件做成小型路由,只負責指出下一步要讀的輔助文件與腳本。模型仍知道去哪裡找完整流程,但不必一開始就載入所有內容。Eric 也提醒,過去常寫成逐步食譜的指示,現在可能把模型綁得太死;保留必要的邊界即可,細節讓模型依任務處理。skill還會影響同一個儲存庫裡其他貢獻者的代理。適合 Sol 或 Luna 的規則,可能對 GPT-6 Astra 太過嚴格;撰寫時要想清楚這些指示會由哪些模型使用。Skill 說明範例:BAD 的適用情境過於籠統;GOOD 明確列出新增或修改資料庫 migration,或審查其部署推行 等觸發條件。上下文預算與 AGENTS.md OpenAI 的 Skills 文件說,初始skill清單最多占模型上下文的 2%;上下文大小未知時,上限是 8,000 個字元。skill太多時,Codex 會先縮短說明,大型集合可能省略部分skill並顯示警告;選定skill後才載入完整 SKILL.md。這是文件描述的載入機制,Eric 的建議則是依這個機制重新整理skill。Eric 建議重新檢查 AGENTS.md:規則應貼近任務,不要每次改一個錯字都要求模型讀完整儲存庫地圖。OpenAI 的 AGENTS.md 文件補充,Codex 會把全域指引與專案層級覆寫依目錄合併,合併後預設上限是 32 KiB。文件連結仍有用,但只留下當下工作需要的內容。測試與完成邊界也要寫得清楚。文章建議在指引中說明本機測試使用可丟棄的測試資料、沒有生產環境存取權,並要求模型修正本次變更造成的失敗後重跑受影響測試。若工作還包含啟動結果、檢查與修正,應在任務提示一開始寫出「完成」包括這些步驟;若希望模型繼續探索,也要寫明探索範圍與停止位置。AGENTS.md 脈絡設定的指引範例圖,上方標示 BAD 顯示籠統的全面讀取指令,下方標示 GOOD 示範依任務需求指派特定檔案的寫法,左下角標註 GPT-6 Astra,右下角標註 Skills & prompting。原文:https://easyvibecoding.app/curated/3275-pvncher-recommends-gpt-6-astra-skill-prompt-simplification
2026-09-06
02 min
REAL ESTATE Strategies with REMAX Hallmark
Why Dylan Provencher Returned to RE/MAX Hallmark
The Toronto real estate market rewards agents who lead with trust…but how do you build it fast enough to survive? Eric Skogland into conversation with Dylan Provencher: team leader, 18-year real estate veteran, and former child actor who bought his first investment property at 17. Dylan shares the unexpected path that brought him from television sets to real estate transactions, and why the referral business he built over nearly two decades traces back to one non-negotiable principle: telling people the truth even when they don't want to hear it. From psychotic cold-calling routines to ha...
2026-08-10
50 min
Accessibility and Gen AI Podcast
Eric Provencher, Founder of Repo Prompt
OUTLINE: 00:00 Opening Teaser 01:26 Introduction 02:39 Eric's Career Path To Launching Repo Prompt 08:01 Making The Decision To Launch Repo Prompt 10:32 What Is A Context Window And Why Does it Matter? 13:38 What Is Repo Prompt Used For? 18:08 How Do You Solicit Feedback From The Users? 21:02 OpenClaw vs Hermes 24:36 Context Engineering vs Prompt Engineering 26:54 Repo Prompt Demonstration and Workflow 45:28 What Have You Learned As An Entreprenuer? 46:38 Why You Should Use Open Source Models with Caution 52:53 Keeping Accessibility In Mind When Coding 01:00:01 Wrap Up -- EPISODE...
2026-04-23
1h 00
Rate Limited
Anthropic Rug pull, WTF is Ralph and more | Ep 8
In this episode of the Rate Limited podcast, hosts Ray Fernando, Adam Larson, and Eric Provencher dive deep into the latest advancements in AI, particularly focusing on Codex 5.2. They discuss user experiences, the steerability and speed of Codex, and the implications of the recent Claude crackdown on third-party agents. The conversation also explores the future of AI in software engineering, the innovative Ralph Loop approach, and the potential disruptions in various industries. The hosts emphasize the importance of adapting to these changes and maintaining a balance between automation and human oversight in coding practices.
2026-01-16
56 min
Rate Limited
2026 Predictions, 2025 Surprises, GPT 5.2 and more | Ep 7
In this episode of the Rate Limited podcast, hosts Ray Fernando, Adam Larson, and Eric Provencher discuss the latest developments in AI, particularly focusing on GPT 5.2 and its implications for coding. They explore the rise of terminal UIs, the evolution of AI coding agents, and make predictions for 2026. The conversation also touches on the surprises of 2025 in AI labs, the potential for consumer AI, and the ongoing debate about whether we are in an AI bubble. The hosts reflect on their experiences with AI tools and share their hopes for the future of technology in software development.
2025-12-19
1h 11
L’Histoire nous le dira
Toute l'histoire du Québec en 40 minutes | L'Histoire nous le dira # 297
C’est quoi l’histoire du Québec ? Adhérez à cette chaîne pour obtenir des avantages : https://www.youtube.com/channel/UCN4TCCaX-gqBNkrUqXdgGRA/join Écriture et réalisation: Laurent Turcot Montage et animation: Martin Bérubé de @proposmontreal Effets sonore: Diane. Artémis Production | artemisproduction.framer.website Vignette: Charles Boidin Merci à Isabelle Picard et Jonathan Lainey pour la révision sur la partie autochtone. Merci à Samuel Rabouin pour ses précieux commentaires sur le script. Pour soutenir la chaîne, au choix: 1. Cliquez sur le bouton...
2025-12-18
40 min
Rate Limited
The Real Cost of Free AI Coding: Episode 2 Rate Limited
SummaryIn this episode of the Rate Limited podcast, hosts Ray Fernando, Adam (GosuCoder), and Eric Provencher dive into the implications of free AI agents, discussing the hidden costs associated with data privacy and sustainability. They explore the performance of Haiku 4.5 compared to Sonnet 4.5, the dynamics of ad targeting in the AI market, and the importance of effective planning and execution in AI models. The conversation also touches on retrieval techniques, the future of AI agents, and the significance of community engagement in navigating the rapidly evolving landscape of AI technology.Takeaways...
2025-10-17
57 min
ThursdAI - The top AI news from the past week
📆 Oct 9, 2025 — Dev Day’s Agent Era, Samsung’s 7M TRM Shock, Ling‑1T at 1T, Grok Video goes NSFW, and Serverless RL arrives
Hey everyone, Alex here 👋We’re deep in the post-reality era now. Between Sora2, the latest waves of video models, and “is-that-person-real” cameos, it’s getting genuinely hard to trust what we see. Case in point: I recorded a short clip with (the real) Sam Altman this week and a bunch of friends thought I faked it with Sora-style tooling. Someone even added a fake Sora watermark just to mess with people. Welcome to 2025.This week’s episode and this write-up focus on a few big arcs we’re all living through at once: OpenAI’s Dev Day and the...
2025-10-10
1h 41
I Build with AI
Is Code a Thing of the Past in Vibe Coding?
In this inaugural episode of I Build with AI, Marcelo sits down with Eric Provencher, founder of Repo Prompt, to unpack the rise of vibe coding—an emerging approach to building software through conversation, intuition, and collaboration with AI. They discuss what separates vibe coding from traditional development, the shifting role of code in the age of large language models, and how AI tools are helping shape the future of AI-assisted creation. Along the way, they explore topics like prompt engineering, tooling for non-developers, and whether we’re witnessing the beginning of a post-code era or just a new way...
2025-08-08
50 min
ThursdAI - The top AI news from the past week
📆 ThursdAI - June 12 - Meta’s $15B ScaleAI Power Play, OpenAI’s o3-pro & 90% Price Drop!
Hey folks, this is Alex, finally back home! This week was full of crazy AI news, both model related but also shifts in the AI landscape and big companies, with Zuck going all in on scale & execu-hiring Alex Wang for a crazy $14B dollars. OpenAI meanwhile, maybe received a new shipment of GPUs? Otherwise, it’s hard to explain how they have dropped the o3 price by 80%, while also shipping o3-pro (in chat and API). Apple was also featured in today’s episode, but more so for the lack of AI news, comp...
2025-06-13
1h 33
The Next Wave - AI and The Future of Technology
The Secret AI Prompt Tool Silicon Valley Engineers Are Using
Episode 57: Can simply "Vibe coding" with AI really replace the need for deep code context when building real applications? Nathan Lands (https://x.com/NathanLands) is joined by Eric Provencher (https://x.com/pvncher), founder of Repo Prompt and an XR engineer at Unity, to reveal the secret AI prompt tool quietly powering Silicon Valley’s top engineers. This episode dives deep into why the current trend of "Vibe coding" with tools like Cursor often falls apart for complex tasks — and how Repo Prompt closes the gap by letting you build effective, highly targeted context for AI codi...
2025-05-06
37 min
Amazon Newsfeed 2022
L'essentiel de RTL - Édition du 12/06/2024
Amazon Newsfeed 2022 - 12/06/2024Au programme :00:00:00 - Jingle 00:00:05 - Le journal RTL: Emanuel Macron, marchés financiers, Hugues Aufray : le journal RTL de 11h du 12 juin 2024 00:03:03 - La météo: Les prévisions météo à 7 jours du 12 juin 2024 00:04:27 - Immersions : les labos du crime: LES LABOS DU CRIME - Affaire Godard : quand les os ont parlé 00:09:45 - Lenglet-Co and You: LES ? DE LA CONSO - Pourquoi le docteur Kellogg a-t-il inventé les corn-flakes ? 00:12:39 - Les invités de RTL: GAUCHE - Raquel Garrido est l'invitée de RTL Bonsoir 00:23:09 - L'œil de Philippe Caverivière: Le 2ème œ...
2024-06-12
2h 20
Dans l'blanc des yeux avec Mélissa Malboeuf
Dans l'blanc des yeux avec Marie-Pierre Provencher
Dans l'blanc des yeux, Mélissa Malboeuf reçoit la fondatrice du mouvement des Multipotentielles, Madame Marie-Pierre Provencher.See omnystudio.com/listener for privacy information.
2023-12-01
31 min
L’Histoire nous le dira
Les sacres au Québec: mystérieuses origines | L’Histoire nous le dira # 222
Au Québec, on sacre! Avec les « torrieu », « mautadit », « viarge » ou encore les « tabarnak », « hostie » et « câlisse », on se retrouve au cœur de l’identité québécoise. Le sacre sert d’exutoire, de marqueur d’intensité ou d’étonnement, de forme transgressive, de signe d’affirmation de soi, d’identification à un groupe, ou encore de rite de passage. Il est un fait social total, en ce qu’il permet de comprendre toute une société. Adhérez à cette chaîne pour obtenir des avantages : https://www.youtube.com/channel/UCN4TCCaX-gqBNkrUqXdgGRA/join Pour soutenir la chaîne, au choix: 1. Cliquez sur le bouton «...
2023-10-30
21 min
Meta Minutes
Meta Minutes (Episode 6 | S3) - Innovative Interactions for the Metaverse
In this episode, we talk to Eric Provencher, a staff research engineer at Unity and a lead developer of Unity’s XR Interaction Toolkit (XRI). We ask him about his views on the metaverse and explore some of the innovative interactions that XRI enables for VR and AR experiences. He explains what XRI is, why it’s useful and how to get started with it12. He also tells us about his passion for natural user interfaces (NUI) and hand tracking for VR/AR. He shares his insights on why hand tracking is so important and whether it will replace VR c...
2023-04-26
25 min
CUCH CAST
CUCH CAST 10
00:00 à 03:12 - Intro parlée 03:13 à 09:54 - CE François Couture - Étude atonale 09:55 à 11:53 - ce qui nous traverse - Qui tu hantes 11:54 à 14:04 - David and the Mountain - Two-minute Electric Toothbrush Cycle, For electric Guitar 14:05 à 17:47 - David and the Woods - Magicien d'os 17:48 à 21:27 - Electrique Junk - Bitchin' Stitchin' (Improvisation) 21:28 à 27:56 - Eric Normand - La lune 27:57 à 32:33 - Guiom & Anette - Suis-je normale (Nini Raviolette) 32:34 à 34:22 - Intervention parlée 34:23 à 36:06 - Marilène Provencher-Leduc - Bout de jaune 36:07 à 38:24 - Pony in the Pancake - Lost I...
2021-12-06
56 min
VR Download
Eric Provencher, Unity Labs Researcher & MRTK-Quest MVP
2021-01-11
28 min
Empouvoirmental
Épisode 13 - Continue ton combat
Nous rencontrons Julie Provencher. Julie est une mère monoparentale de 46 ans qui a vécu de la violence conjugale et qui, grâce à sa famille et ses enfants, reprend aujourd’hui sa vie en main. Pour du soutien, de l’information et des références concernant la violence conjugale, vous pouvez contacter SOS violence conjugale de différentes façons :Par téléphone : 1 800 363-9010 (24h/7) Par courriel : sos@sosviolenceconjugale.ca (délai possible de 24h) Par texto : 438-601-1211 (entre 14h30 et 22h, du lundi au vendredi) Site web : http://www.sosviolence...
2020-10-04
25 min