[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] #  @melissapan Melissa Pan Melissa Pan posts on X about ibm, intesa sanpaolo, prompt engineering, ai the most. They currently have XXXXX followers and XX posts still getting attention that total XXXXXX engagements in the last XX hours. ### Engagements: XXXXXX [#](/creator/twitter::1715115161806073856/interactions)  - X Week XXXXXXX +2,366% - X Month XXXXXXX +41,341% - X Months XXXXXXX +559% ### Mentions: X [#](/creator/twitter::1715115161806073856/posts_active)  - X Week X +60% - X Month XX +500% - X Months XX +220% ### Followers: XXXXX [#](/creator/twitter::1715115161806073856/followers)  - X Week XXXXX +13% - X Month XXXXX +22% - X Months XXXXX +166% ### CreatorRank: XXXXXXX [#](/creator/twitter::1715115161806073856/influencer_rank)  ### Social Influence **Social category influence** [stocks](/list/stocks) XX% [technology brands](/list/technology-brands) XX% **Social topic influence** [ibm](/topic/ibm) 20%, [intesa sanpaolo](/topic/intesa-sanpaolo) 20%, [prompt engineering](/topic/prompt-engineering) 10%, [ai](/topic/ai) 10%, [link](/topic/link) 10%, [mins](/topic/mins) 10%, [secret](/topic/secret) 10%, [year of](/topic/year-of) 10%, [gain](/topic/gain) XX% **Top accounts mentioned or mentioned by** [@neuripsconf](/creator/undefined) [@nobanksnearby](/creator/undefined) **Top assets mentioned** [IBM (IBM)](/topic/ibm) [INTESA SANPAOLO (ISNPY)](/topic/intesa-sanpaolo) ### Top Social Posts Top posts by engagements in the last XX hours "(5/N) ➕ RQ2 - How are production agents actually built We discover that simplicity and controllability win 🏆 💬70% of teams use off-the-shelf frontier models with no finetuning relying on prompting 👷79% of deployed agents lean heavily on manual prompt engineering (agent prompts can be 10k tokens) 🪡Workflows are short and controlled: XX% of agents run XX steps before a human steps in XX% X 🔧85% of case studies skip third-party agent frameworks and build custom apps In practice teams deliberately limit autonomy to keep agents reliable" [X Link](https://x.com/melissapan/status/1996984852332400683) 2025-12-05T16:47Z 2865 followers, 1451 engagements "we are in front of door E Aisle Data-Centric AI. Starting in 10min" [X Link](https://x.com/melissapan/status/1996653559186927800) 2025-12-04T18:51Z 2770 followers, XXX engagements "🚀 Building agents in real-world production Wed love to talk Were running a large-scale study on production agents - a collaborative effort from UC Berkeley IBM Stanford UIUC Intesa Sanpaolo and others. Join our study today I promise it only takes X mins Link in 🧵" [X Link](https://x.com/melissapan/status/1961077562345415005) 2025-08-28T14:45Z 2838 followers, XXX engagements "Its happening today 11am-2pm at Exhibit Hall CDE Poster XXX Come by our poster Ill tell you everything about MAST (and maybe our secret project too) 🤔" [X Link](https://x.com/melissapan/status/1996595683256643747) 2025-12-04T15:01Z 2864 followers, 13.5K engagements "Thrilled to release our new paper MAP: Measuring Agents in Production ⚙🚀 2025 is the year of agents but do they actually work in the real world Is it just hype A group of XX researchers from Berkeley Stanford UIUC IBM and Intesa Sanpaolo investigated what makes agents deployable in the wild. So 📈 Why agents Productivity gains ➕ How to build production agents Simple & controllable methods 🧑💻 How to evaluate agents Heavy human oversight 🛑 Top challenge now Reliability remains unsolved We surveyed XXX agent builders and ran XX in-depth interviews across XX agent application domains to" [X Link](https://x.com/melissapan/status/1996975916971626763) 2025-12-05T16:12Z 2865 followers, 118.4K engagements "(4/N) 📈 RQ1 - Why agents The primary usage of agents now is to augment human users for productivity and efficiency gain. XX% of the agents are built for human users Signal potentially heavy human oversights. 👓 We also observe an interesting trend of focus on latency lenient use cases 🐌: XX% of deployments tolerate response times of minutes or more while only XX% require sub-minute latencybecause even slow agents still beat human baselines. So AI agents are already running in production and creating real-world impact across XX diverse domains" [X Link](https://x.com/melissapan/status/1996979811718504482) 2025-12-05T16:27Z 2865 followers, 2448 engagements "(6/N) 🧑💻RQ3 - how to evaluate agents in development humans are *still* heavily involved in the loop. 👀 XX% of deployed agents rely primarily on human-in-the-loop evaluation 🤖 XX% use LLM-as-a-judge but every in-depth case study teams that does this also adds human verification Public benchmarks rarely fit bespoke production tasks. 📊 XX% of teams build custom benchmarks dataset (with gold labels) XX% skip benchmark datasets entirely and lean on A/B tests expert review and user feedback Evaluation in the wild is a challenge domain-specific and still very human" [X Link](https://x.com/melissapan/status/1996985696062754918) 2025-12-05T16:51Z 2865 followers, 1374 engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@melissapan Melissa Pan
Melissa Pan posts on X about ibm, intesa sanpaolo, prompt engineering, ai the most. They currently have XXXXX followers and XX posts still getting attention that total XXXXXX engagements in the last XX hours.
Engagements: XXXXXX #
- X Week XXXXXXX +2,366%
- X Month XXXXXXX +41,341%
- X Months XXXXXXX +559%
Mentions: X #
- X Week X +60%
- X Month XX +500%
- X Months XX +220%
Followers: XXXXX #
- X Week XXXXX +13%
- X Month XXXXX +22%
- X Months XXXXX +166%
CreatorRank: XXXXXXX #
Social Influence
Social category influence stocks XX% technology brands XX%
Social topic influence ibm 20%, intesa sanpaolo 20%, prompt engineering 10%, ai 10%, link 10%, mins 10%, secret 10%, year of 10%, gain XX%
Top accounts mentioned or mentioned by @neuripsconf @nobanksnearby
Top assets mentioned IBM (IBM) INTESA SANPAOLO (ISNPY)
Top Social Posts
Top posts by engagements in the last XX hours
"(5/N) ➕ RQ2 - How are production agents actually built We discover that simplicity and controllability win 🏆 💬70% of teams use off-the-shelf frontier models with no finetuning relying on prompting 👷79% of deployed agents lean heavily on manual prompt engineering (agent prompts can be 10k tokens) 🪡Workflows are short and controlled: XX% of agents run XX steps before a human steps in XX% X 🔧85% of case studies skip third-party agent frameworks and build custom apps In practice teams deliberately limit autonomy to keep agents reliable"
X Link 2025-12-05T16:47Z 2865 followers, 1451 engagements
"we are in front of door E Aisle Data-Centric AI. Starting in 10min"
X Link 2025-12-04T18:51Z 2770 followers, XXX engagements
"🚀 Building agents in real-world production Wed love to talk Were running a large-scale study on production agents - a collaborative effort from UC Berkeley IBM Stanford UIUC Intesa Sanpaolo and others. Join our study today I promise it only takes X mins Link in 🧵"
X Link 2025-08-28T14:45Z 2838 followers, XXX engagements
"Its happening today 11am-2pm at Exhibit Hall CDE Poster XXX Come by our poster Ill tell you everything about MAST (and maybe our secret project too) 🤔"
X Link 2025-12-04T15:01Z 2864 followers, 13.5K engagements
"Thrilled to release our new paper MAP: Measuring Agents in Production ⚙🚀 2025 is the year of agents but do they actually work in the real world Is it just hype A group of XX researchers from Berkeley Stanford UIUC IBM and Intesa Sanpaolo investigated what makes agents deployable in the wild. So 📈 Why agents Productivity gains ➕ How to build production agents Simple & controllable methods 🧑💻 How to evaluate agents Heavy human oversight 🛑 Top challenge now Reliability remains unsolved We surveyed XXX agent builders and ran XX in-depth interviews across XX agent application domains to"
X Link 2025-12-05T16:12Z 2865 followers, 118.4K engagements
"(4/N) 📈 RQ1 - Why agents The primary usage of agents now is to augment human users for productivity and efficiency gain. XX% of the agents are built for human users Signal potentially heavy human oversights. 👓 We also observe an interesting trend of focus on latency lenient use cases 🐌: XX% of deployments tolerate response times of minutes or more while only XX% require sub-minute latencybecause even slow agents still beat human baselines. So AI agents are already running in production and creating real-world impact across XX diverse domains"
X Link 2025-12-05T16:27Z 2865 followers, 2448 engagements
"(6/N) 🧑💻RQ3 - how to evaluate agents in development humans are still heavily involved in the loop. 👀 XX% of deployed agents rely primarily on human-in-the-loop evaluation 🤖 XX% use LLM-as-a-judge but every in-depth case study teams that does this also adds human verification Public benchmarks rarely fit bespoke production tasks. 📊 XX% of teams build custom benchmarks dataset (with gold labels) XX% skip benchmark datasets entirely and lean on A/B tests expert review and user feedback Evaluation in the wild is a challenge domain-specific and still very human"
X Link 2025-12-05T16:51Z 2865 followers, 1374 engagements