Kenyan with love for STEM | Innovation | Finance | Development | Africa | Travel || Tweets are my own opinion. Retweets and tweets are not endorsements||
1 in 5 UK visitor visa applications is refused — and the £135 fee is non-refundable.
See how ready your application looks against the actual UK visa rules before you pay. Free, under 10 minutes.
Built for Nigeria · Pakistan · Ghana · Kenya · Bangladesh
We hired a new VP of Engineering who is obsessed with agile methodology.
He called a meeting on his first day and said we need to transition to 2-week development sprints.
He wanted daily stand-ups, retrospective boards, and continuous deployment pipelines.
He wanted us to actually write new code.
I realized immediately that he was an existential threat to my lifestyle.
I let him finish his impassioned speech about workflow velocity.
Then I stood up, walked to the whiteboard, and drew a single horizontal line.
I told him agile sprints are a localized solution for a localized mindset.
I said our infrastructure operates on a Zenith Release Cycle.
He asked what a Zenith Release Cycle was.
I told him it's a holistic, macro-stabilization framework where we observe the system in a state of prolonged stasis.
By not touching the code for 18 months, we allow the legacy dependencies to organically settle.
I told him that deploying bi-weekly updates creates micro-abrasions in our database architecture.
I used the phrase chronological data scarring.
The CEO was in the room and audibly gasped.
He told the new VP that we can't risk chronological data scarring just to satisfy a trendy tech buzzword.
The VP looked at me like I'd just invented a new color.
He was completely paralyzed by the sheer density of my fabricated jargon.
He quietly agreed to adopt the Zenith Release Cycle.
We're officially scheduled to deploy our next update in the third quarter of 2027.
I spent the rest of the afternoon buying things I don't need on Amazon.
Agile is a disease invented by people who want to be punished for their salary.
I refuse to participate in my own suffering.
When Shohei Ohtani was a high school freshman, he created a detailed "dream sheet" with one central goal: to be the #1 draft pick for 8 NPB (Nippon Professional Baseball) teams.
It was a 64-cell roadmap based on a framework called the Harada Method.
Here's exactly what Shohei did 👇
1. First, some history.... The Harada Method was created by Takashi Harada, a Japanese junior high track coach. He took a team ranked last out of 380 schools and, using his system, turned them into the #1 team in the region within 3 years. They held that top spot for the next 6 years.
2. You start by placing your main goal in the center of an 8x8 grid. For Ohtani, this was "be the #1 draft pick."
3. Next, you identify 8 critical supporting pillars needed to achieve that goal. These surround the main goal.
Ohtani's 8 pillars were:
• Body
• Control
• Sharpness
• Speed
• Pitch Variance
• Personality
• Karma/Luck
• Mental Toughness
4. You then break down each of those 8 pillars into 8 smaller, actionable tasks or daily routines.
This fills out the entire 64-cell grid, turning a massive dream into a concrete, daily action plan.
To improve his karma, he listed tangible actions like:
• Showing Respect to Umpires
• Picking up trash
• Being positive
• Being someone people want to support
5. The method goes far deeper than just technical skills. It forces you to analyze your weaknesses and build confidence. It also has a highlight on service to others, emphasizing that humility and contributing to your community are essential for personal success.
6. The key to the system is daily execution and accountability. Once the 64-cell chart is complete, you turn the tasks and habits into a daily diary and a "Routine Check Sheet." It’s designed to transform abstract intentions into a measurable, daily practice.
icymi we wrote a new agents book: patterns for building ai agents
it has everything you need to take your agents from prototype to production, like agent design patterns, the basics of security, etc
reply to this tweet with BOOK and we'll dm you so you can get a copy
It’s fascinating how quick people are to mention names like KJ and Jaguar whenever my leadership journey comes up yet they rarely acknowledge visionaries like my brother @BobiWine, with whom I’ve worked on projects across Kenya and Uganda, or global icons like George Weah and other creatives who turned their art into instruments of transformation. Leadership isn’t a comparison contest, it’s a calling born from lived experience, awareness, and a sense of purpose. Our journeys may differ, but our mission for change is one. Kibera will rise, not by chance, but by choice, by unity, and by vision. The future is local.
#KibraRising
Today’s AI slop was yesterday’s food.
Before algorithms ate our words, farmers fed leftovers to pigs. It was a closed loop. What people couldn’t use became nourishment for something else. Nothing was wasted because nothing could be.
Then industry found shortcuts. Distillers fed whiskey mash to cows. Their milk turned blue. Vendors hid the rot with plaster of Paris. Thousands of infants died.
When everything speeds up, discernment is the first casualty. Meaning follows close behind.
Now the same hunger for speed drives our machines. AI trains on recycled data. Feeds learn to feed themselves. Each cycle thinner, faster, emptier.
We’ve mistaken velocity for vision. We call the residue progress and pretend it still has flavor.
Rot only wins when we stop noticing the taste.
The farmer feeding scraps to pigs knew the difference between sustenance and waste. The distiller didn’t.
Our choice is the same.
And history’s warning is sitting right there in the trough.
I got my latest stool test (GI-MAP) exactly a month ago and after much testing on all AI models, my conclusion is this: They just sort of suck at diet and gut health.
Here are a few problems I've noted:
1: Some models (notably ChatGPT) just default to standard-of-care and FDA approved "safe" recommendations. They generally shy away from integrative and functional medicine, as well as most focused diet and nutrition guidelines. They also tend to keep entirely separate the clinical literature (i.e. what a doctor would recommend) with what the actual science says. For instance, if you ask framed in a medical context like "would kidney beans help me here?" the answer is generally "no" but then if you ask "what does the science say?" the picture changes DRAMATICALLY
Note: to be fair, this is just the state of the industry. Most doctors know zilch about nutrition, which is just a commentary on the sorry state of medicine. Even gastroenterologists know startling little about microbiomes and diet.
2: They generally struggle to think systematically by default. However, if you ask for things like RCA (root cause analysis) or other specific terminology like "characterize pathological phenotype" they are much better. But your average user doesn't know to ask for an RCA or pathological phenotype.
3: Their recommended interventions are not bad, but they are wildly variant across chatbot models and even across conversations. Initial prompting is extremely important, even when all chats and models are given the same root clinical data. This is concerning because, if there was a good understanding in these models, they SHOULD generally come to the same or similar conclusions. (to be fair, when they are all extremely well prompted, they do tend to converge). But without good prompting and background knowledge (which most users possess neither) it's a crapshoot on the recommendations.
I think that this comes down to a few things:
FIRST - the training data. Simply put, gut health and nutrition are simply not represented well enough in the data. All the pieces are there, but it's like they are completely siloed. Chatbots tend to have a "doctor mode" and a "scientist mode" and it's like they rarely overlap. Even when you prod them to try and reevaluate clinical views with what has been scientifically established, it's like talking to a doctor who's ideas are 20 years out of date.
SECOND - the training schemas and UX. A user should not have to be an expert in something already to get passably useful advice. When I first started using AI to evaluate my GI-MAP it basically gave me a shrug like "you might have a handful of minor problems" but when prompted correctly it was like "no, this is actually kinda serious and here's a full breakdown." The fact that it CAN synthesize across silos (but only when directly prompted) or perform RCA is interesting. Shows that the underlying technology is likely capable, but post-training needs to be better.
Attached, an example of a prompt that got decent, reliable, repeatable results across models.
Note: my ChatGPT style tells it to use complete sentences and paragraphs only, and to limit output to 2 or 3 paragraphs. This prevents endless rambling.
If you have 32 minutes, I will show you how Nairobi County loses 32,949,346,981.
Yes, that's not a typo. KSH 32 billion.
Let me walk you through a trip to hell.
Data from 2023-2024.
The county had KSH 33 billion to spend. To serve 4.6 million residents.
The county employes around 16,000 people. 0.3% of the county population.
Step #1 - First order of business: Use 55% of the budget to pay salaries, to this 0.3% of the population.
That is KSH 18 billion. Now you have KSH 15 billion left.
You are told that the county entered into contracts with certain firms in 2021. The contracts were all supposed to lapse in 2022. They lapse. There are no extensions.
KSH 6.7 billion.
Side Note: Ghost workers galore. Quoting the auditor:
"Review of the bank remittance for April, May and June, 2024 revealed that seven thousand, seven hundred and seventy-seven (7,777), six thousand, one hundred twenty-three (6,123) and six thousand, eight hundred and three (6,803) officers respectively, shared the same bank accounts, agent code and branch code."
So let’s understand this fact pattern.
This year’s wage bill goes up by 55% relative to last year. And the city did not hire 55% more people or raise salaries by 55%.
Then, you find out that, the auditor found out that between 6,123 and 7,777 people who got paid by the county as employees, shared their bank account with someone else who also got paid.
In other words, between 6,123 and 7,777 of the county’s 16,000 employees are couples.
That is what Sakaja wants you to believe.
In addition, quoting the auditor verbatim:
“Review of the bank remittance for August, 2023, revealed that seventy-four (74) officers shared the same name.”
74 officers have the same biographical information.
Step #2. Use KSH 1.2 billion to traverse the globe for totally useless trips.
In a city full of garbage - as you can see in one of the attachments, the county politicians and their lackies pull out their passports.
Book flights to Morocco - Marrakech, Morocco - a resort city, at the cost of KSH 37 million. They want to go study "proactive management". In Morocco. Two full weeks!
Look at examples:
19 people — Marrakesh, Morocco — “Proactive Management Programme — Ksh 37,235,843 (15 days, 10th–24th June 2024)
8 people — United Arab Emirates — “Conference Facilities and Training”— Ksh 29,771,200 (7 days, 12th–18th Feb 2024)
14 people — Vancouver, Yale Town, Canada — “Seminar on Leadership Management and Conflict Resolution” — Ksh 10,920,000
9 people — Italy — “Training on Strategic Stewardship” — Ksh 9,822,120 (8 days, 13th–20th Dec 2023)
14 people — Canada — “Workshop on Leadership Management & Conflict Resolution"— Ksh 9,611,748 (7 days, 5th–11th Dec 2023)
6.7 people — Dubai — “Personal Branding and People Professional Summit” — Ksh 8,993,450 (16 days, 18th Aug–2nd Sept 2023)
8 people — New York — 8th United Nations General Assembly — Ksh 8,782,764 (8 days, 16th–23rd Sept 2023)
9 people — Dubai, UAE —" Esami Pre-Retirement and Pension Planning Programme”— Ksh 8,132,061 (8 days, 29th Apr–6th May 2024)
8 people — Vancouver, Canada — Conference — Ksh 6,761,520 (17 days, 22nd July–7th Aug 2023)
6 people — Istanbul, Turkey — Budgeting for Performance Management Programme — Ksh 6,751,523 (11 days, 12th–22nd Oct 2023)
6 people — Singapore — Financial Analyst Planning and Control Programmes — Ksh 6,735,227 (10 days, 1st–10th Nov 2023)
6 people — Turkey — Budgetary Performance Management Course — Ksh 6,699,198 (11 days, 12th–22nd Nov 2023)
The city is full of garbage. And @sakaja and his lackies are flying from Marrakesh to Vancouver.
Proactive management. Conflict resolution. Personal branding. Retirement.
You could make up this stuff, even if you wanted to.
When they are done with these shenanigans, Ksh. 1.2 billion is gone.
You have to wonder: Is there no course on how to collect garbage? Nothing like that, even in Galilee Israel?
Folks, this is insanity. This is not the behaviors of people who have common sense and business sense. It is not the behaviors of people who treat every shilling as the last shilling we have. Which is what must happen in government affairs.
So, KSH 1.2 billion is wasted like that.
Total damage so far? KSH 6.7 billion in salary heist 1.2 billion in wasteful travel=KSH 7.9 billion.
Let’s keep going.
KSH 839,933,289 in salary expenses cannot be substantiated because there is ZERO records. No names of employees, departments, evidence of payment. Just a claim. Basically, KSH 839 million in the KSH 18 billion I mentioned above have no support whatsoever. This wage bill itself increased by 55% this year compared to last year.
The county still employes less than 0.5% of the county population.
Step #3. Lose Kshs.140,944,662 through a “duplicate payment” to a “vendor”. No explanation for why a duplicate payment was made. And this is not corrected.
Step #4. Lose Kshs.478,485,041 to pending bills that nobody has ever heard about.
Step #5. Lose another Kshs. 2,005,937,441 in relation to pending bills.
Pay attention here.
The county says it owes KPLC Kshs.724,964,162, but when auditors verify the balance with KPLC, they find out that the county actually owes Kshs.2,730,901,603.
This likely says that, the county pretended to be paying down pending bills at some point, but the funds went somewhere else, such as, a corrupt official’s bank account. So the bill is still outstanding, but the cash in the city’s accounts is gone.
Step #5. Lose another Kshs. 791,013,848 in relation to pending bills. The auditor finds an additional 791 million in bills that are owed, but which the county has not reported.
At this point, let’s make sure you understand how pending bills come about.
If you are a government and you are smart and accurate in budgeting, you can estimate what you will spend. And if you are Nairobi County and you have Ksh 33 billion to spend, you have enough to pay your way. But, if in budgeting you put down “Ksh 500 million in luxury travel and then spend Ksh 1.2 billion, then you won’t have Ksh 700 million to pay the entire amount. That is a bill that ends up being hidden, or increases pending bills. But there are also many cases where corrupt politicians introduce pending bills as a way to steal. You blame the previous government, and then prioritize paying new claims that have no support.
Or, cash is stolen from bank. That cash was designed to pay certain bills, and now, there is no money to cover those bills. They become "pending".
Step #7. Lose Kshs.6,269,546,657 to 4 lawyers. Of the actual “genuine pending bills”, is Kshs.6,269,546,657 owed to just 4 lawyers. Equivalent to 20% of the annual budget. Can you believe that?
Listen to the auditor:
"It was noted that most of the Court cases related to issues such as disputes of unpaid claims for goods/works/services completed by contractors, un-procedural termination of employment contracts, irregular procurement processes and poor contract management. The judgements entered against the County Executive resulted to a high cost of litigations and interests."
In other words, corruption and mismanagement of county affairs take another Kshs. 6.3 billion. 100% wasteful.
Step #8. Lose another Kshs.1,393,234,865 in compensation that cannot be supported. Payment vouchers generally used for temporary workers worth 1,393,234,865 have no support. No one can tell you John Maina or Alice Omondi was employed here, for x amount of time, and got paid Ksh X.
Step #9. Lose another Kshs. 28,861,597 in compensation that cannot be supported.
Within the healthcare delivery expenses, it is found out that (62) National Identity numbers were used to pay people duplicate payments. In other words, one person appears in sone system used to pay. They get paid. You look at another system used to pay people for different things. The same person, same ID, appears again, and is paid.
Consider these additional damning factors:
>>Some officers’ basic salary changed multiple times during the year with some changing 6 times
>> Three thousand, two hundred and sixteen (3,216) officers changed job groups more than once, twenty-four (24) of whom changed job groups three (3) times.
>> Three (3) officers had their payroll numbers attached to different National Identification numbers at various times of the year.
>>Twenty-four (24) officers had their payroll numbers attached to different Tax pins at various times of the year.
>> One hundred ninety-seven (197) officers had their birth dates changed multiple times.
Step #10. Lose another Kshs. Kshs.301,400,000 in “bursaries” that cannot be accounted for. The county says we sent money to schools X, Y, and Z.
The auditor tries confirming these claims. And nobody can confirm it.
Step #11. Lose Kshs.140,313,872 in claimed “civil works expenses”.
You are told that the county entered into contracts with certain firms in 2021. The contracts were all supposed to lapse in 2022. They laspse. There are no extensions.
There is NOT ONE document telling you what the contractors were to do, and for how much.
But the county goes ahead and pays Kshs.140,313,872.
Step #12. Lose Kshs.32,964,416 in a contract to build a kitchen. A contract is irregularly issued to a firm to construct a kitchen.
The auditor says:
“Physical verification of Central Kitchen for the programme in October, 2024 revealed several issues at Toi Primary Kitchen. Visible cracks were observed on the wall, despite the kitchen being completed only a year earlier.
There was no electricity meter installed. In addition, water connectivity was not installed by the contractor, instead, the implementor connected the water at their own cost. There was also poor workmanship of the pavements and the installed gas pipe was not connected to the cooking vessels, rendering it unusable.”
Step #13. Lose Kshs.10,103,488,814 in a contract to build “affordable” homes.
Any contract exceeding Ksh. 5 billion requires approval by the Attorney General. The county won’t get it. They use “specially permitted procurement” method where only one bidder provides a quote. So there is no competition, which is likely why they did not seek the Attorney General’s approval.
@SakajaJohnson and his lackies select the desired partner. ZERO competition. They don't seek the approval required.
A note on this particular one: There are actually 4 contracts valued at Ksh 31 billion which the auditor is expressing doubts. About legitimacy of the amount.
Quoting the auditor:
"In the circumstances, the ownership of the land, value for money and regularity for the housing projects contracts totalling Kshs.31,047,436,654 could not be confirmed. In addition, Management was in breach of the law"
My estimated loss is the Ksh 10 billion as presentative of what the loss could be. It is the only "estimate" in this analysis, and again, it is less than the Ksh 31 billion the auditor is questioning.
Step #14. Lose Kshs. 2,282,899,172 in stalled projects.
Quoting the auditor:
"Review of construction projects undertaken by the County Executive revealed that projects valued at Kshs.2,282,899,172 in respect to the construction of various projects have stalled. It was not clear why these projects remained stalled for years while the County Government embarked on new projects, some that still end up stalling due to non-payment of contract sums due. Failure to manage the projects in line with law may lead to loss of public funds and poor service delivery."
Step #15. Lose Kshs. 16,510,760 by awarding contracts to the highest bidder, when lower ones are available, and against the law.
Step #16. Lose Kshs.223,304,547 in shoddy works and projects. The County issues 4 contracts to build 4 stadiums at the cost of Kshs.1,036,551,255.
Stadiums in Kawangware, Ziwani, Kangemi and Dandora.
After Kshs.223,304,547 is paid out, the auditor notes:
>>The floodlights were non-functional and the internal lighting had not been installed.
>>The field’s surface was uneven, with noticeable bumps and depressions in certain areas, which rendered it unsuitable for football activities.
>>The electrical cables and wiring system on the field were exposed, increasing the risk of theft and vandalism.
>>The roofing materials used were substandard and showed visible signs of deterioration due to exposure to direct sunlight. According to the field management, rainwater frequently leaked through the roof, leading to water accumulation in the seating areas during rain.
>>The public address system was non-functional, as explained by the Management, who confirmed that the speakers were not operational.
>>There were visible holes in the perimeter wall. Rather than allowing footballs to bounce back into the field upon contact, footballs created holes due to penetration.
The stadium walls were constructed of soft boards, which could easily be penetrated with little force instead of durable concrete stone walls.
Total damage: KSH 32,949,346,981
@Mizani254@MoGAbdi@PropesaTV@HillaryGondi@fit_ermined@EricLatiff@MigunaMiguna@NairobiCityGov@KenyaGovernors
If dictators in Uganda, Kenya and Tanzania have united against the people of East Africa, surely the people of East Africa SHOULD equally unite against them.