ThatTechBro (🛠,🚀)

If you are a designer (product designer, ux designer, ui designer, ui/ux designer) and needs a resume written to pass the ATS screening, you can reach out to me. It will cost you a token 🙂

“I will start blocking fools like you” as though you are paying me to tweet. Mtchwww

@TosinOlugbenga @GammsApp And you resorted to insults? Really?

@404secnotfound You are very daft. Someone gave himself a challenge to build a replica and you said i didn’t mention what? He doesn’t know what @GammsApp can do before attempting. I will start blocking fools like you now

You should have mentioned all of these in your initial tweet but you didn’t. You are trying to deflect and find loopholes not to hold up your own end of the bargain. Just let the lad know you don’t have the capacity to hold up your own end of the deal and keep it moving

Is there anyone in @OpenRouter team I could talk to?

Mastered 13+ programming languages, started tech in 2022. Huh? *head scratch*

This Rhema’s discuss is revealing so many things of which some are laughable: - It reveals that so many people do not know how the tech ecosystem is. - It reveals that everyone who claim to love/like open source don’t actually like it or understand what open source is or how it works. - It reveals that Nigerians see everything as competition rather than compliment Everyone saying “until it’s your idea.” Ideas are worth nothing, you are not the only one with a working brain. Building has never been the hardest part in startups neither are ideas. The hardest is still distribution and with the age of AI, you would need the 7 gods to help you. I understand the sentiment of seeing someone else build out what you have spent so much time building after you launch, what they can’t take away from you is your process. Keep building.

Getting to a week I started running the third phase of my three week experiment after decoupling Minion and the only issues I noticed are: - Ability to translate what I want to how the agent understands the task - this happened during setup, so its a one time occurrence in this experiment. - Cron job firing keeps alternating between 1 hour and 2 hours even though its set up to run every 1 hour, I'll chalk this up to network issues due to bad weather we had earlier in the week

Due to network interruptions, cron job is now running every 2 hours instead of every 1 hour. So far, all outputs are following same pattern - no deviation

I’m dragging myself along trying to do the transition of Zeroshot from v1 to v2 and its speed is stressing me out. Trying to push something that won’t break the users’ experience when I could have just yanked the whole thing out

@asemota @osuide Really interesting in not so fun ways. Since it’s relatively a new launch, let’s keep watching the trajectory

I asked Claude about Mythos and became even more worried. cc: @osuide, @404secnotfound This is very fresh news — announced just yesterday (April 7, 2026). Here's a summary: Claude Mythos Preview is Anthropic's most powerful model to date — described internally as a "step change" beyond even Opus. Here's what we know: What it isAnthropic describes it as "by far the most powerful AI model we've ever developed," with a new tier called "Capybara" — larger and more capable than Opus — that Mythos appears to be built on. Why is it being withheld from public release? During testing, Anthropic found that Mythos Preview is capable of identifying and exploiting zero-day vulnerabilities in every major operating system and every major web browser. The exploits it constructs are sophisticated — in one case, it chained together four vulnerabilities into a complex browser exploit that escaped both renderer and OS sandboxes. How capable is it, really? Opus 4.6 turned Firefox vulnerabilities into working exploits only 2 times out of several hundred attempts. Mythos Preview, running the same benchmark, produced 181 working exploits and achieved register control 29 more times. How it was tested: Anthropic deployed it via Claude Code in isolated containers, prompting it to simply "find a security vulnerability in this program." It would read code, hypothesize vulnerabilities, run the project, add debug logic, and output bug reports with proof-of-concept exploits. Project Glasswing. Rather than a public release, Anthropic launched Project Glasswing — an effort to give organizations that build or maintain critical digital infrastructure early access to Mythos Preview to find and patch vulnerabilities before the model (or similar ones) becomes broadly available. An important nuance: These cybersecurity capabilities were not explicitly trained into the model — they emerged as a downstream consequence of general improvements in code, reasoning, and autonomy. It's a significant moment. Anthropic is essentially saying this model crosses a threshold where responsible release requires a coordinated defensive deployment first. Given your interest in AI infrastructure and agentic platforms, this is directly relevant — the security layer of agent infrastructure just became a much more urgent problem.

The gap between a human’s instruction and the interpretation of said instruction by an LLM or model is still wide. Before you talk about prompt/context engineering or skill issues, I shouldn’t write a dissertation for an “intelligent” being to understand what I’m trying to convey. I understand some people have issue articulating their thoughts, this is not that. This is a gap that should be resolved

Set a reminder for my upcoming Space! x.com/i/spaces/1rGmq…

We will do this at 6 pm GMT or 7 pm Nigerian time today. Please join us as we discuss all that we have been doing or trying to do with AI. Hopefully, my bosses will join us too. @kenju254 @emekaokoye @MafiaSlice @perfexcellent @404secnotfound @hackSultan @osuide @kidehen

It’s a new week to either conquer or be conquered, always choose the former and not the latter. 🥂

@asemota @MafiaSlice @perfexcellent @kenju254 @EmekaOkoye @kidehen @osuide Sure

@MafiaSlice, @perfexcellent, @kenju254 @EmekaOkoye @404secnotfound can we do this tomorrow afternoon or evening? cc: @kidehen @osuide

“Summarize what we have done so far” Current session: 18% What the hell @bcherny @AnthropicAI

I went to church today for the first time since 7 years and it was worth it - why I love going to Catholic Church. I also realized that I have forgotten some of the prayers and the reason why I was always agitated whenever the prayers are sang rather than recited - no be today that agitation start.

From prompt engineering to context engineering, now harness engineering? The hats next on the menu?

I’m in the process of reworking Minion because the last three weeks of run makes me feel like it’s been over engineered

I believe we are over complicating agents’ architecture especially when it comes to certain attributes like memory.