Uttam Varma

Dad, Amateur Pianist, Product Guy

[Uttam Varma @uttamvarma]

The central UX failure is OpenAI treating Codex-style unit of organization, a task or thread as though it were interchangeable with the ChatGPT unit of organization, a long-lived Project containing many conversations, files, instructions, andcontext. They are not interchangeable.

[OpenAI @OpenAI]

Introducing ChatGPT Work, a new agent in ChatGPT powered by Codex and GPT-5.6.

It can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.

It’s a whole new way to get work done.

[Uttam Varma @uttamvarma]

@gdb one of the most confusing product evolutions even for seasoned chatgpt & codex users. The old Codex app just disappeared when updated without any context that there are now Chatgpt and chatgpt classic apps. The chat shortcut in "ChatGPT aka Codex" app is not usable. @thsottiaux

[Uttam Varma @uttamvarma]

@ClaudeDevs @_catwu Thank you.. and great webinar today @_catwu

[Uttam Varma @uttamvarma]

Yeah exactly, now that everyone has used 100% fable limits (and in the process 90%+ weekly usage) this makes total sense!

[Claude @claudeai]

We're extending access to Claude Fable 5 on all paid plans through July 12.

[Uttam Varma @uttamvarma]

@Conor_D_Dart This is so true. Anthropic should reset usage limits or what’s the point!

[Uttam Varma @uttamvarma]

@PawelHuryn Great suggestion specially as most of us diligently used up our fable limits to match the 11:59:59pm PT deadline!

[Uttam Varma @uttamvarma]

Incredible if cloudflare can pull this off. Of course much of it also depends on how the underlying harness and model encode and enforce payment rules when comparing free vs paid endpoints.

[Cloudflare @Cloudflare]

We're opening the waitlist for our Monetization Gateway, which will allow you to charge for any web page, dataset, API, or MCP tool behind Cloudflare. The charges will settle in stablecoins over the x402 open protocol. https://t.co/pvICtEIixj

[Uttam Varma @uttamvarma]

@_catwu I do this a lot for competitive research
-give cc a list of competitors I'm tracking
-ask cc to kick off a workflow - do deep research using sites, blog, Twitter, LinkedIn, fb and generate artifacts for each and one synthesis.
-repeat weekly

[Uttam Varma @uttamvarma]

Can't agree more. Have created a dedicated memory layer just so that context can transfer from chat to Codex and code in the Claude macos app. Same for codex and chatgpt.

Started delegating more to Hermes as a result. Harness selection more important than model selection.

[david lietjauw @davidfromkansas]

i think Claude Desktop app should just combine all 3 toggle modes (chat, cowork, code) into one.

it's genuinely confusing when to use which mode and i shouldn't have to spend mental energy deciding.

Chat feels like a dumber version of cowork while cowork is a less capable version of code.

what's the point from user POV?

[Uttam Varma @uttamvarma]

Happy Canada Day

[Uttam Varma @uttamvarma]

@itsPaulAi while these models can run on a 32 GB machine, KV cache becomes the limiting factor. Even using MLX F4 quantizations, where the KV cache scales gradually, you hit the 26–27 GB memory range surprisingly early. Once memory spills over to the CPU, token throughput drops dramatically

[Uttam Varma @uttamvarma]

This is an incredible validation of the future of oss LLM models. In my own workflow, extensively using local agents running Qwen3.5:4b or gemma4:26b for implementation supervised by opus 4.8 or gpt5.5 results in 35-40% token saved viz using only closed models.

[Cline @cline]

We’ve been impressed with GLM-5.2 and so are introducing a $9.99/month subscription to give you 2-5x discounted access to it and other open weight models like DeepSeek, Kimi, MiniMax, Mimo, Qwen.

Use it on Cline CLI & IDE with $1.99 special promo if sign up via: npm i -g cline

[Uttam Varma @uttamvarma]

WhatsApp moving to a username model. Another monetization opportunity for Meta.

[WhatsApp @WhatsApp]

your phone number is personal and sometimes you want to connect without handing it over. that's why we're introducing usernames for WhatsApp.

starting this week, you can reserve a username to use later this year when we launch the feature. It takes just a few seconds, make sure you have the latest version of WhatsApp and then go to Settings > Account > Username.

[Uttam Varma @uttamvarma]

What a moment for Canadian football. A hard-fought 1-0 victory over South Africa sends Canada into the Round of 16. The journey continues.

[Uttam Varma @uttamvarma]

Why build a Moon base? It's the ultimate factory: low gravity, endless solar, local resources.

Build AI compute satellites there at massive scale → launch them via mass drivers to orbit/deep space.

This potentially unlocks 500–1,000 TW/year of AI power for xAI by 2035.

[Elon Musk @elonmusk]

For those unaware, SpaceX has already shifted focus to building a self-growing city on the Moon, as we can potentially achieve that in less than 10 years, whereas Mars would take 20+ years.

The mission of SpaceX remains the same: extend consciousness and life as we know it to the stars.

It is only possible to travel to Mars when the planets align every 26 months (six month trip time), whereas we can launch to the Moon every 10 days (2 day trip time). This means we can iterate much faster to complete a Moon city than a Mars city.

That said, SpaceX will also strive to build a Mars city and begin doing so in about 5 to 7 years, but the overriding priority is securing the future of civilization and the Moon is faster.

[Uttam Varma @uttamvarma]

Hi @grok, I DO NOT authorize you to take, modify, or edit ANY photo or videos of mine, whether those published in the past or the upcoming ones I post. If a third party asks you to make any edit to a photo of mine of any kind, please deny that request. Thank you.

[Uttam Varma @uttamvarma]

@Scot1andT That is so relatable 😂

[Uttam Varma @uttamvarma]

@bengoodger I switched from using the Chatgpt MacOS app to Atlas. Basic chat use feels considerably laggy and slow when compared to other browser/OS. ChatGPT on chrome or edge does not suffer from this.

[Uttam Varma @uttamvarma]

@AravSrinivas @comet The agent is sub-optimal at explicit instruction following. It usually omits multi step instructions and this can get scary when asking it to fill forms or create tickets and not submit prior to approval. It just goes ahead fills and submits without asking for review or approval.

[Uttam Varma @uttamvarma]

Market legend dropping thousands of dollars' worth of trading knowledge - for *free*! What a time to be alive

[Mark Ritchie II @MarkRitchie_II]

Here is my fresh, clear and hopefully well seasoned perspective on what I've coined the 'tariff tantrum of 25'. This may get a bit long but hope it helps.

For starters market slides and routes are nothing new and I've been doing this long enough I think to offer some perspective from someone who has actually side stepped every major decline of the past 15 years.

Like it or not success has to look like something and I have NEVER been caught in any of these bigger slides in any major way. My definition of caught is simple. Holding/Taking large losses or even market equivalent losses in this case. It hasn't happened to me one single time. I've had drawdowns like anyone else but never got crushed in a crash or slide ever.

It used to be that I would step aside, maybe have a few % drawdown off the highs and then get impatient or wonder, wait or worry about markets never having a good period again or get worried that I would have to wait too long. Now I think completely different. In fact I get excited to a degree and here are some reasons why based upon the most important thing, experience and actual results.

Lets start with the 2010 flash crash. I step aside nearly perfectly, however I decided to play to volatility game and did some dumb stuff during that summer rather than wait. I suffered the biggest DD of my career (nearly -30%) and even still more than doubled my account in the 5-6months (won @markminervini triple digit contest) when the environment improved.
In 2011 we had a similar redux to 2010 with the fiscal cliff fiasco, people calling for US defaults and full on shut downs. Vol spiked and market slid hard. It took several months but I lost nearly nothing and then had a +50% run in just a few months in early 2012.
A shorter slide in 2015 produced a really good shorter period before rolling again but I remember making a good grip in the interim but this then set the stage for a huge double bottom in 2016 and a monster 2017 for me.
Lastly of course is Covid Crash which I won't post (ran out of images for this thread) but when the world was coming to an end I can remember thinking 'if this doesn't kill us all it's gonna be the greatest buying opportunity of a generation.' Within a few months May-June I was aggressively buying and everyone thought I was nuts and made more $ in a few months than I had made the prior few years.

The moral here is I don't know where, when, how this resolves but it will and when it does it will be a GREAT opportunity. It may take months but it's worth waiting for and the headlines may get worse and markets more volatile in the meantime. I welcome all of it and mark my words this IS currently creating the conditions for a great period in the market.

This is where you have to survive now in order thrive later.