original title: how non-developmental teams work like engineers (without writing code)...
Photo by Damien Player
Photo by Peggy Block Beats

Compiler presses: When most people still see AI as a "more efficient search tool", Perplexity is starting to work。

THE PAPER REVOLVES AROUND A GAP THAT HAS BEEN REPEATEDLY IGNORED — WHY AI IS BEING USED THE SAME WAY, SOME GET ONLY ONE PIECE OF THE ANSWER AND OTHERS GET THE DELIVERABLES DIRECTLY. THE KEY IS NOT TO MODEL CAPACITY, BUT RATHER TO USE IT AS A DIALOGUE WINDOW OR AS AN IMPLEMENTATION SYSTEM THAT CAN BE DIRECTED AND MOVED。

A new type of tool, represented by the Personality Computer, uses the word "task" instead of "question" as a core interactive approach. From contract clearance, competitive analysis to data cleansing and reporting generation, users no longer describe problems but define the final delivery directly. This ability is further evolving from a one-time output to a reusable, self-operable workflow, in line with the linking of business tools and the consolidation of personal background and style models。

More importantly, automated boundaries are being redefined. It is no longer merely a complement to a certain step, but can be sustained, cross-tooled and even offer additional tasks. This means that the relationship between people and tools is moving from "use" to "management and entrustment"。

IN THIS CHANGE, THE REAL WATERSHED IS NO LONGER WHETHER AI IS BEING USED, BUT WHETHER IT IS ALREADY BEING USED AS A DELIVERY RESULT。

The following is the original text:

Those who understand this will gain an asymmetric advantage. Soon, everyone will learn how to do it. But before everything becomes obvious, here's the way you can start early。

Over the past year, developers have been running autonomous AI intelligent bodies (e.g. Claude Code, OpenClaw, etc.) that can do their own research, build products, deliver the full results directly, and do not need to be watched over and over again. But you don't really use that -- unless you use the terminal, you write the code。

And Perplexity Company changed that. This is the first time that non-developers can use the same capabilities. You just need a browser and a task that you can give it。

Most people open Perplexity, enter a question, get an answer, and then close the page. They missed the key. It's not for answering questions, it's for carrying out missions。

Stop asking questions and start giving it the real job。

Why would most people fail

The finance director, the lawyer, the consultant... they opened the tool, entered a question, got a pretty good answer, and then they thought, "Oh, a higher Google." Then they went on to spend 90 minutes cleaning up the form that was cleaned up last Monday。

The problem is not the tool, but the use. They used it as a chat robot。

Question method:"What are the risks of this contract?"

Mission approach:"Examination of the contract. An article-by-article verification of open-source support for all statements; marking of vague wording, missing provisions and possible legal liability; listing of the five most critical risk points, with reference to specific articles; and the output of a Word document with modified marks

The same contract. One way is to give you a list and you can read it yourself; the other way is to give you a copy of the product that can be distributed to the client。

It'll only take 10 minutes to get this system up

Connect the tools first. Click on the connetors of the sidebar. Perplexity can connect to over 400 applications: Gmail, Google Drive, Slack, Salesforce, Notion, SharePoint。

THEN LET IT KNOW WHO YOU ARE. "I'M IN A POSITION TO WORK IN SOME KIND OF COMPANY." I'LL PRODUCE X, Y, Z ON A REGULAR BASIS. REMEMBER THESE BACKGROUNDS DURING EACH SESSION. IT WILL KEEP THIS INFORMATION FOR A LONG TIME。

And tell it what's good. "These are my best work examples." Learn their format and tone, which will be used as a reference for subsequent content generation.”

So it's not guessing your style, it's dismantling the path you've tested。

Ten minutes, do this first。

A real example: a Monday that no longer takes 90 minutes

A financial analyst receives a data export once a week, 150 rows, with a confusing format: duplicate data, three date formats, ratings written in words rather than numbers. Before starting analysis, she spends 90 minutes a week cleaning data. The same issue, repeated weekly。

She entered only one instruction: to clean the file, to remove it, to standardize the date format, to convert the text rating into a number; to analyse the cleaned data; to generate an interactive dashboard with filters and to provide a shared link; to output a PDF report before and after a comparative cleansing; and to save all files to the "Monday Report" folder of Drive。

4 Minutes Later: Clean data sets, interactive dashboards, shared links, PDF reports - all in her Drive。

Then she asked, "Did I not ask, but would it be more useful?"

The system recommended two points: first, that the task be automatically run at 7 a.m. a week; and second, that a new task be created to generate a Tuesday management briefing based on a poorly performing panel。

She set both up and turned off the page。

And then every Monday, it automatically runs -- whether her computer is on or off。

This is the set of capabilities that developers have been using over the past year. Now, you can use it in the browser。

People are already using it for what

@gregisenberg did a live test on @startupidespod。

He was given only one task: to find the company that advertised the competition, to identify the real sponsors and to write a personalized e-mail for everyone。

The system found Ramp ' s growth vice president, grabbed the content of a podcast he had participated in two weeks ago, wrote a cold e-mail, quoted his specific statement in the programme and sent it directly. Greg did not say "send" and the system judged the mission to be completed and self-executed。

It then took the initiative to monitor the competition podcasts, alerting and attaching the corresponding contact, "Call as soon as the budget starts."。

Eventually, the process completed a parallel survey of 96 potential clients and arranged a 3rd and 7th day follow-up mail。

The team used it to audit the entire HubSpot product page: auto-crawling the entire site, scoring by custom criteria, sorting questions and generating a shared web report. It would have taken a team to work for a week, which was done when the programme was recorded。

It's all done on site, not a demonstration, not a preset script。

Work-specific usage

YesFinancial areaAn investment portfolio analyst gave only one assignment before the release of the Ingweida newspaper。

The result of the return was a real-time interactive dashboard consisting of $13.5 billion in revenue, 75 per cent Māori rate, 114.2 per cent growth rate, full profit statement, and a projected profit trend from fiscal year 2021 to 2028, all supporting filtering and sharing links。

No Excel, no manual search data, 5 minutes done。

Perplexity has direct access to data sources such as SEC Disclosure, FactSet, PitchBook - no API key, no additional authorization, and complete in the system。

Legal scene:
"Examination of the contract. An article-by-article verification of open source support for all formulations; the marking of vague wording, missing standard terms and elements of possible legal liability under [state-specific] contract law; the listing of the five most critical risk points with reference to specific articles; and the output of a Word document with a modified mark

One of the reviewers had uploaded a proposal claiming that a market had increased by 43 per cent over the same period. The real data was only 4%, and the problem was stopped before the signing。

Marketing scene:
"Analysing [competing 1], [competing 2], [competing 3] the best performing content over the past 30 days; identifying the most interactive content forms and themes; identifying content gaps; generating a 30-day content calendar based on these gaps and preserving it as Google Doc."

Set it as a time job. The latest competition analysis is automatically generated on Monday, without manual research。

Operating scene:
"THIS IS OUR Q1 CSV DATA. PLEASE CLEAN UP THE DATA; ANALYSE THE INCOME BY REGION AND PRODUCT LINE; IDENTIFY THE THREE BIGGEST PROBLEMS; GENERATE ONE PAGE OF PROPOSALS FOR ACTION; PRODUCE ONE PAGE OF REPORTING PPT; AND SAVE ALL DOCUMENTS TO THE PROJECT FOLDER

Five deliveries, one directive. It was finished when you met。

Model Evaluation: 60 seconds to three judgments

When you face a decision with practical consequences, you only have to enter the question once. Perplexity will call Claude, ChatGPT and Gemini at the same time, and will be summed up by a "complex" of their consensus and disagreement。

• Convergence of the three: high faith conclusions

Part of the difference: further judgement

Some ask if the price of the product is $297 or $497. The three models give different answers, but the only unanimous conclusion of the compositers is that they are not below $297. Decision-making is complete。

Many companies would pay consulting firms to hold analysts in conference rooms to draw conclusions。

Here, just one order。

Real core competencies

In order to get real value from Perplexity Computer, 80 percent depends on one thing: can you clearly describe the final output。

Not technology configuration. It's whether you know what you're going to deliver. Do not describe steps, do describe results。

After each mission, remember to ask: "Did I not ask, but would this result be more useful?"

It almost always points to the blind zone. Every time。

Start here

Open Perplexity (pro version $20/month). Enter the Computer page, click on conners, to connect Gmail and Google Drive first。

Enter your three sentences of background introduction (only once). Upload 2-3 of your best work examples so it can learn your style. And then pick a task that you spent more than two hours last week and each output was similar: describe it in a "final delivery" way, send it. Observation of the implementation process. If you repeat a task, set to run automatically in front of the closed page。

The developers have been using this for a year. The difference between their output and that of others is real。

This is the way to close the gap。

[ Chuckles ]Original Link]