The best advice I'm ever going to give you about software development is this:
Don't start with the tech or the logic - begin with the people. Software is about people. Build a solution that solves a real problem, and the rest will come together.
It will be interesting to know when (and if) this massive change in jobs will hit - will may continue to be slow enough and mixed into other change so it’s not seen or felt.
Companies are already replacing their headcount with tokens. Big companies with layoff rhetoric will brand and blend it into other decisions.
It will be interesting who survives the blast radius when AI-assisted tools drop that businesses can adopt and implement with ease.
Startups pitching. AI isn’t a category, it’s the norm with 60% as AI-native, 26% is AI-enabled, and only 14% has no AI core features.
The defining theme is “AI employees,” not copilots, not assistants. The pitch is always “we replace [expensive human role] end-to-end,”
A milestone for @TWiT as it turns 21! (In the US, that means I could legally have a drink with it!)
This milestone brings back a memory of being invited by @leolaporte to join the live stream in the studio, to talk about the tech industry in #Rwanda.
I still remember when podcasts weren’t “podcasts” yet. I was digging for Apple/Mac content and randomly landed on @leolaporte first podcast episode. (it was his only one-up at that time - accidental early adopter status unlocked)
I’ve been listening to TWiT shows pretty much every week since, watching it grow from scrappy livestreams into one of the most consistent voices in tech media.
Years later, after meeting @alexlindsay in Rwanda a few times, he said, "When you are next in California, come by the studio.”
And, I did.
Leo, being Leo, just waves me over and drops me into the big table conversation… live streaming. No prep. No warning. Just: “You’re on!”
We talked about tech in #Rwanda and the rise of the software industry in #Kigali.
It was fun - and a cool opportunity to be on the TWIT network!
@LucyMbabazi I only started to hear about it after I traveled to Africa and started living in Rwanda. It’s used in numerous podcast to show the regional difference as economic strength of the “north” vs “global south.” I don’t think it’s an attack, but does help keep the gap apparent.
I can’t send a simple message to my mother or my wife anymore without wondering what kind of accidental crime I just committed. 🫨😖😅
A couple days ago, I tried to tell my mom I loved spending time (“handing out”) with her a few months ago when i was back in the states. I wrote that I look forward to when i get back so we can ‘hang out’ again.
However, my iPhone autocorrect wrote, “I really want to hang you.” 😵🤣
This was in a family group chat, and there was a different family dynamic between us that day - and a lot of teasing. Luckily, I am my mother’s favorite and i’m still in good standing, but my iPhone is not helping me keep that status.
I’ve been writing on mobile devices since the late 90s. I type rapid because my brain is far ahead of my typing. My wife calls it ‘my intense typing.’ Her teasing is somewhat true, as I get a lot of information in my head, and start to type fiercely to get it all down. This leads to mistakes and the need for restructuring sentences. Somehow, I am worse on an iPhone today than I was 10 years ago. Not because I forgot how to type, but because the keyboard has decided it’s a co-author - and often not a good one!
I don’t think I’m crazy or my skills are worsening. And, to prove it, I did the research - I found, it’s not just me…
It’s was easy to find a huge increase of other people with similar complaints - articles are everywhere with titles like, ‘the keyboard got “smarter”… and forgot its one job.’ Across Reddit and other social media, there are long rants. Searching on google and AI Perplexity, clearly describes there is an upswing in frustrated mobile auto-correct use. It’s a thing.
Others are running into the same thing - missed taps at high speed, words getting rewritten after you type them, autocorrect confidently “fixing” things that were already correct. And massive complaints on missing the context (like hanging your mother).
It was reassuring to know that many people have experienced auto-correct altering the context of a sentence, such as changing “hanging with my mom” to “hanging her,” with a delay that makes it harder to notice. This is a common issue many have encountered.
What is really silly here, is that AI would write any sentence preferring that someone would rather ‘hang their mother’ vs ‘hanging with their mother.’ Obviously we’re are still in the early stages of AI being helpful, (And more so for all the moms out there. 😜)
Under the hood, it makes sense: More machine learning, more personalization. The keyboard is literally learning how you type. The problem is… sometimes it learns the wrong lessons.
A few typos, some slang, one weird phrase, and suddenly your phone starts overriding perfectly normal sentences like it knows better. Add a bit of lag, a few dropped inputs, and now you’re fighting both the keyboard and its opinion about your writing. My style of writing often is a bit snarky, mixed in with the serious or direct tone = which i’m sure is hard for AI predict.
That’s the real shift.
It’s not just more typos. It’s a system that misinterprets what you wrote, and then aggressively commits to being wrong. And the results of all this new technology helping me write better (this is does), but also can twist it into the wrong context.
So, if you’ve typed a message lately and immediately found it was auto-correcting out of context - you’re not along.
Lastly, this was written on a whim, on a Saturday morning, in a lovely UK cafe next to the river Themes, completely in response after some good natured family teasing of my DM typing failures. So, if you find any typos or out of context sentences or meanings, it wasnt me who may not be in professional mode, or focused to write a professional or quality post, and i’m surly not the one that poorly wrote anything - it was the iPhone marking all the mistakes (can i use that as an excuse from more on? Maybe if that is good and logical, i can use it to explain that last poorly written DM to my mom 😅😉)
Rwanda just did something most ecosystems are still debating - It built a Kinyarwanda ICT dictionary with 1,700+ tech terms! 🙌🏻
This is great! We are all about Rwandans building tech solutions for Rwanda at @RwandaBuild - If people can’t understand technology in their own language, they can’t fully participate in building it.
Rwanda Develops First Kinyarwanda ICT Dictionary
Rwanda has reached a milestone in digital inclusion with its first Kinyarwanda ICT dictionary, compiling over 1,700 tech terms. Developed by the Rwanda Cultural Heritage Academy with partners, it standardises digital vocabulary to make technology more accessible. By translating complex concepts into Kinyarwanda, it challenges the idea that African languages cannot keep pace with innovation.
The dictionary reflects a wider push to preserve linguistic identity in the digital age. Supported by the Rwanda Education Board, it will strengthen education, support AI development, and reduce reliance on foreign languages. As Africa’s tech ecosystem grows, such efforts highlight the need for inclusive, locally rooted digital systems.
So, in the last two months, @AnthropicAI shipped 75 major features and modules on Claude (Claude — Code, Cowork, Dispatch, computer control, 1M context on desktop, scheduled tasks, more)
That’s not just “fast execution” or “aggressive cadence.”
That’s not a roadmap, that’s a firehose of capability directly into the core product that devoted core users already use every day - that’s a different level of dev operating and focus entirely.
Their users can’t stop talking about how awesome all these tools are - they are getting all the buzz and feel like the clear leader for many months now.
Then, take their competitor and leader in the AI space, @OpenAI, which is shutting down Sora, its text‑to‑video/video‑generation product, just months after launching it to major fanfare and signing a high‑profile Disney licensing deal.
While Anthropic accelerates efforts to empower individual Claude users, OpenAI is shutting down popular products to focus on and expand enterprise and partner channels. Clearly, OpenAI is developing and likely working on a major strategy away from direct end-user focus, targeting what partners will utilize for their end users.
- Anthropic has 1,000+ employees, Claude + safety + infra; “AI‑native” product org
- OpenAI has 3,500+ employees, ChatGPT, an enterprise API, tools, robotics, and AGI research
Only time will tell which strategy will pay out - maybe both strategies earn each to share future pieces of the AI pie.
For Women Who Give Everything, This Night Is For You.
Join a powerful evening designed for women across Kigali - whether you're in tech, business, healthcare, school, or building your life in between.
Limited spots - secure yours now https://t.co/EHq3YTKgzY
“AI solutions as advanced agents will run every SME in the business side in the years to come.” - @rickard
Hoover, small business owners don’t know how to build useful AI agents; no time to learn agentic workflows.
New/Better solutions for operators are needed in the market!
Great products aren’t built from bursts of effort.
They come from sustained, focused execution.
If you can’t focus on one task for hours at a time, you won’t build anything meaningful.
If you can’t stay locked on one problem for days, you won’t build anything great.
Get to work
@CommonWorld ...@ThibautMntwari & @willynsabiyumva, you two see and talk to a lot of founders and entrepreneurs in #Kigali - what do you see and hear on seeking to hire top talent that is maybe 'too good' for the current stage of the business?
I also find that in Africa, hiring higher-paid talent for leadership roles in a startup is more like hiring a mercenary for the short-term battles you are in. The culture is more of a problem-solver than a long-term growth or leadership culture, which causes a loss of culture and vision.