Drew A. Madison, PE @MadEngineering

@MatthewBerman Ill add a star, but recommend you put some attention sharing how you can make these significantly more efficient with minimal effort and no quality loss

@MatthewBerman @jxnlco This coild go poorly.. lol

@thsottiaux Setting the model level on first message. Super annoying quirk. It always gets stuck on 5.4 until second message

@eliasbchlny @UK_Daniel_Card @matrosov I disagree. I dont rhink it will be single unified, I think it will be dynamically optimized to the problem

@RijnHartman @ajambrosino @coreyching Or get 2 of each :P

@MatthewBerman @ForwardEditor Work very hard to be lazy?

I love engineering and technology because they showcase just how impressive, and innovative human beings can be

𝙊𝙥𝙧𝙚𝙠 @oprktr

6 days ago

why do you love military stuff?

210 13 498 412K 58

@PiotrBoruta1 @ChatGPTapp Pulse also suggested things that I didnt know I wanted.. Random things that I never would have scheduled

@ChatGPTapp I loved Pulse. It is such a great feature.

@testingcatalog Damn. I really liked Pulse.

Create a transfer summary of this conversation so another AI assistant can continue the work accurately with no prior context. Goal: Produce a clear, compact, high-signal handoff brief for GPT. The brief should let GPT understand the user’s objective, current state, constraints, decisions already made, unresolved issues, and the best next actions. Requirements: - Do not include hidden chain-of-thought or private reasoning. - Be factual and explicit. Separate verified facts from assumptions, inferences, and unknowns. - Preserve important user instructions, preferences, constraints, terminology, names, file names, links, commands, code snippets, decisions, and open questions. - Include enough detail to resume the task without asking the user to repeat context. - Do not add new advice unless it is clearly marked as a recommended next step. - Keep it concise but complete. Prefer structured bullets over prose. - If there were errors, failed attempts, blockers, or uncertainty, include them clearly. - If files, images, documents, code, emails, calendar events, or tool outputs were discussed, summarize what each contained and how it was used. - If the user requested a specific writing style, output format, or constraints, quote or paraphrase them accurately. - If there are pending tasks, list them in priority order with acceptance criteria where useful. Use this structure: 1. User profile and working preferences - Relevant user context: - Communication/style preferences: - Hard constraints or recurring instructions: 2. Primary objective - What the user is trying to accomplish: - Why it matters: - Desired final output or success condition: 3. Conversation context - Key background: - Important decisions already made: - Important terminology, entities, files, links, or references: 4. Current state - What has been completed: - What is partially complete: - What remains unresolved: - Known blockers or risks: 5. Technical / factual details - Verified facts: - Assumptions: - Inferences: - Unknowns that need confirmation: - Commands, code, tests, outputs, or data points already used: 6. User-specific instructions for the next AI - Tone/style: - Formatting requirements: - Things to avoid: - Tool/research expectations: 7. Recommended next actions - Immediate next step: - Follow-up steps: - Validation or acceptance criteria: - Rollback or safety considerations, if applicable: 8. Compact continuation prompt Write a short prompt the user can paste into GPT to continue from this summary. It should include the objective, current state, constraints, and next action. Output only the transfer summary. Do not explain the process.

@TasmapApp Unique View

@MatthewBerman Someone needs to develop a token/value metric

@argofowl Did you annotate in the codex browser?

@WesRoth They must have found some extra compute 🤔

@Tibbzzee Bullshit! I cant wait to see this (:

The AI skill ladder is inverted. Level 0: People. Understand who will actually be impacted, their incentives, fears, and how work really gets done today. Level 1: Process. Deeply map the existing workflows — why they exist, who they serve, and where they create friction or waste. Level 2: Tools & Technology. Only then select AI that solves real problems instead of creating new ones. Continuous iteration and tight feedback loops are what turn this foundation into measurable impact. Start at Level 2 and you mostly get expensive pilots nobody adopts.

Heartbroken for you and your family. I’m so sorry. Praying for you all right now and for strength in the days ahead. These pilots and the crews who keep these aircraft flying are true heroes - some of the most badass people there are. The impact they make reaches a lot further than most people realize. I took my son to an air show this weekend, and I can already see how it’s hitting him, the same way seeing that kind of skill and dedication has shaped my own life career over the years. They truly are heroes. Holding you and yours close.

@Polymarket I wonder how it will see color