Daniele Tassone

@hellonehha @AnthropicAI Last week Jasmine caught response degradation on a legal edge case. One customer affected. Fixed in 1h. The report is what is actionable — Jasmine itself is supervised, not autonomous (deliberate decision). It analyses clearly enough that the next step is obvious/actionable!

@danieletassone @AnthropicAI +1 on observability. When you say Jasmine is agent observability actionable - can you explain what all action it takes once a agent fails?

After reading @AnthropicAI blog on Agentic AI. spent some time to create a mental model to understand how to design, and explain Agentic AI architecture Define a task/goal - what you want agent to do achieve? 1. Orchestration layer : it is your control panel 3. Agents layer: this layers made of agents (multi /specialised) 4. tools: your tools are made of this layer (web search, DB, APIs etc) 5. memory: this is the brain to store information - long or short term etc. 6. monitoring : This is the most crucial to monitor each and every step 7. Reliability & failure management: identify errors, retry, fallback, involve human 8. Governance and security: compliance, audit, auth etc.

Yesterday we dropped Genie 3.0. -> Works at unlimited context length for complex legal transactions -> Provides tabular insights across all your past documents or deals -> Outperforms Claude & GPT for legal tasks -> Gives you a chat + doc editor view together. Let us know what you think! #legalAI @GenieAI

Genie 3.0 is live. Most legal AI loses the plot by document four. Genie 3.0 holds the whole matter in its head and cites every clause. Eidetic intelligence under the hood. 140% more accurate than ChatGPT on legal. Reply "access" to try free 🚀 #genieai #legalai #aiagents

"I can behave like a lawyer." Rahul Ranjan, Sales Engineer at HoSt Bioenergy. Hundred-page contracts across Europe and North America. Reviews run 80% faster with @GenieAI. New HoSt Group case study out today.

@PawelHuryn @karpathy @PawelHuryn Thank you so much, this is really valuable work 🙌. I confirm, I came to similar conclusions a few days ago using approaches B and C for a PoC I'm working on. I love people who use a scientific approach to validate hypotheses.

After an interview with @karpathy, everyone is talking about what AI agents can/can't do. But an opinion without data is just a hypothesis. So, I tested 3x185 workflow executions for a market researcher agent. The results have shocked me🧵

The first Football player signing under @GenieAI sponsorship!

"Hardening sprints" (a bug-fix phase) are waterfall thinking at it's worse. As Deming said: "Inspection is too late. The quality, good or bad, is already in the product." Why is the quality that your "hardening Sprint" adds not already in the product? Fix that.

Delivering "on time and cost" is an anti-pattern. Delivering garbage cheaply and quickly benefits nobody. Cancelling a project that's not going anywhere is not a bad thing. The key is to work in very small increments so that you can cancel as early as possible, based on feedback.

"Agile Practices are 268% More Likely To Fail"... WHAT A LOAD OF... (link to the video in my bio) #agile #softwareengineering

Creating a cache layer for a database query that takes 9 seconds will sure give the client a sub millisecond latency. But that is just a duct tape. Blindly and completely replacing the database on which the query takes 9 seconds for a database that takes 3 seconds, is just a more expensive duct tape. It is a matter of time until the new database shows its own problems. Understanding why the query takes 9 seconds in the first place and fixing it is the most effective path forward. Even if the decision ends up replacing the database, at least now you know why exactly you are replacing the component, it is not a black box to you. This understanding is not easy because tech is truly overwhelming. It takes time, effort and dedication to understand. I’m not advocating to never change infrastructure. On the contrary, by understanding how languages, frameworks, databases and tools work, we may make informed design choices to replace the tool because of fundamental deficiencies. Understand the system. There are no shortcuts. You can’t get rich quick.

With a friend like Genie AI, you're the boss when it comes to legal documentation 🧞 But how does it work? 👇 Check out the demo below and see the magic in action 👇 #artificialintelligence #ai #legal #law #legaltech #lawtech #technology

How to spot an A-Player: -> Independent Thinker -> Curious & Humble -> High Standards -> Energy Raiser -> Focused Context matters. But character matters more.

There's a difference between (frequent) deployment and (infrequent) releases. You need fast feedback for agility. You don't, however, need to release to the entire user community to get that, though it's good if you can.

📢 Exciting news! 🤖✨ We plan to outperform GPT-4’s accuracy on specific legal tasks by year end! How? Massive and proprietary datasets, additional LLMs and prompt engineering hand-in-hand with leading UK lawyers. Get ready for a game-changer in legal tech! #AI #LegalAssistant

Recently, we at @GenieAI open sourced a Clause Library with the aim of helping #legal clauses standardization 📚 by using @MongoDB Atlas Search. Atlas Search was a great choice: ⏩ Impressive search results 🕹 Easy to use 😁 Easy to learn read more: genieai.co/blog/legal-cla…

Valve is one of the most agile companies I know of. No Sprints. No backlogs. No frameworks. No Scrum. No SAFe. Just pure agility built around iterative development and feedback. It's worth 15 minutes of your life to see how they work: youtu.be/9Yomqk0C6kE 1/2

@Franc0Fernand0 Deal with people is a great skill. People build components and these components have to deal with other components. If we can't deal with people, how can these components work?

Coding is only a small part of software development. In order to advance, you must also: • Think critically • Solve problems • Deal with people • Design architectures • Know CS fundamentals It goes beyond programming languages and frameworks.

@housecor Great point and well summarised! Based on my experience, I have observed that developers tend to be more focused on finding solutions rather than understanding "the problem" (alias "the context").

Some developers love consistency too much. Examples: “Always use let.” “Always use an arrow function instead and of a function declaration.” “Always extract an event handler, even if it’s not reused and a single line.” Avoid “always”. Consider the context.

@allenholub Well said Allen ✌🏻 We don’t have sprints where I’m working at the moment. It requires good communication and collaboration between team members 🗣️ but done properly it's a huge win 😎

<sigh/> Let's just get rid of the damn sprints entirely! When you're done with the current thing, pull the next thing. The entire team works on one story at a time. It takes as long as it takes.