manishsharan

9 ms·

91.
by manishsharan 2y ago
I too was puzzled by the response from Claude. I am using the Anthropic workbench with claude-3-5-sonnet-20241022 (latest) But it think it has to do more with the freshness of training data. AWS IPV6 Egress is a new technology from AWS wh
92.
by manishsharan 2y ago
Leaderboards are not that useful for measuring real-life effectiveness of the models atleast in my day-today usage. I am currently struggling to diagnose an ipv6 mis-configuration in my enormous aws cloudformation yaml code. I gave the same
93.
by manishsharan 2y ago
>> hard to deny that their LLM models aren't really, really good though. The context window of Gemini 1.5 pro is incredibly large and it retains the memory of things in the middle of the window well. It is quite a game changer fo
94.
by manishsharan 2y ago
On a related topic, I tried to use Selenium with Chrome with Google search. My idea was to find the result , read the search result web pages and feed it to an LLM and have it summarize the information. Somehow Google detected it was a bot
95.
by manishsharan 2y ago
I think the reason some of us continue using Google Analytics is its demographic data. That information is not available elsewhere as far as I know , which I admit is not a lot.
96.
by manishsharan 2y ago
In case of an accident, who will be held liable? Tesla Inc. or the poor sod driving remotely?
97.
by manishsharan 2y ago
I have not .. but this seems to be something I must try.
98.
by manishsharan 2y ago
Here is how I am implementing something close to what you mentioned. In my setup, I make sure to create a readme.md at the root of every folder which is a document for me as well as LLM that tells me what is inside the folder and how it is
99.
by manishsharan 2y ago
That's why everyone in North America drives or aspires to drive a Ford F150 or Hummer
100.
by manishsharan 2y ago
I use it to help drive engagement with the product for paying subscribers to introduce features that they have not used and invite them to training webinars.
101.
by manishsharan 2y ago
This. I have been searching for SES on this discussion and nobody seems to be using it except perhaps you and I. My email marketing costs are a fraction of all the providers listed in the article thanks to AWS SES. And building a small cust
102.
by manishsharan 2y ago
Hold on.. In my experience I found that ChatGPT writes awesome Clojure code. So much so that most of my clojure code in the last few months were written by clojure. sure it gets some stuff wrong but overall it knows more clojure functions t
103.
by manishsharan 2y ago
>>but really this is also a wake up call for any programmer like me who is overly dependent on a single technology such as dotnet or java Java is open with competing implementations from Oracle, Redhat,AWS and of course OpenJDK and se
104.
by manishsharan 2y ago
>> If you knew this ahead of time then what value did it provide you? I think I should have clarified that the responses were accurate when we bench-marked them against RAG applications built on other LLMs. If the results had been po
105.
by manishsharan 2y ago
I agree with you. For business applications, context window is more important than any extra reasoning ability. I can load up nearly all business documents, regulations and rules and ask it questions and the responses are accurate and bette
106.
by manishsharan 2y ago
I have found Gemini pro to be excellent for RAG application. Most of my java projects with less than 500k lines of code can fit entirely in its context window so I don't have to mess with chunking. Before this came along, we had tried
107.
by manishsharan 2y ago
If only we could get rid of the "Return to Office" being pushed by corporate , this could turn out to be great for young families.
108.
by manishsharan 2y ago
Svelte vs React is discussion that will never end. For those of us who do not like to code complex logic and state in javascript, I can recommed htmx https://htmx.org/ . Also, I am more comfortable writing backend code and t
109.
by manishsharan 2y ago
I , for one, thank the investors for subsidizing AI and making it affordable for us. I am not sure my measly US$20 per month will be enough to cover the cost of my usage but I am grateful for the fun I am having while the music plays.
110.
by manishsharan 2y ago
Gemini is also very cheap and comparable and works well as a drop in replacement for OpenAI models.
111.
by manishsharan 2y ago
No. This is the same logic that I used when I did not buy Amazon and missed that ride. The market is not always rational.
112.
by manishsharan 2y ago
I do not have experience with AWS RDS so I cannot speak to that. In my experience, GCP Cloud SQL for Postgres has been more expensive than MS Azure SQL. In our tests, CloudSQL also was not comparable to the resiliency offered by Azure SQL.
113.
by manishsharan 2y ago
MS Azure SQL on is very cheap compared to any other self hosted database including PG and MS SQL. Unless they running a super heavy workload , this solution will meet most business requirements. But this is also a gateway drug for clo
114.
by manishsharan 2y ago
Glad to know I am not the only parent who feels this way. On a personal note, I am one of those people who highlights sections of text books using different coloured markers and scribbled notes on the margins. During my final exam time, my
115.
by manishsharan 2y ago
I have been meaning to ask the observability experts this question: Why not dump all metrics , events and logs into Clickhouse ? and purge data as necessary? For small to medium sized businesses/solution ecosystem, will this be be enou
116.
by manishsharan 2y ago
>> What large corporation would ever sign up to that? Speaking of the banking sector, we are migrating our workloads to the public cloud as it allows us to be nimble and responsive to business needs. And we are getting a highly resil
117.
by manishsharan 2y ago
Not having to wait for procurement to approve to buy a few terabytes of NAS is nice. Not dealing with procurement cycle for h/w purchase is priceless.
118.
by manishsharan 2y ago
Several famines. https://en.wikipedia.org/wiki/Timeline_of_major_famines_in_I... .
119.
by manishsharan 2y ago
Gemini Advanced seems very promising with 1 Million token context window. https://one.google.com/explore-plan/gemini-advanced I haven't tried it as I use Claude primarily.
120.
by manishsharan 2y ago
I have found chatgpt to be shockingly good for writing functions. I decompose my problem domain into a series of functions and I have chatgpt generate code for each those functions. My greatest fear is that I am losing my memory when in com

More ›